BPG is committed to discovery and dissemination of knowledge
Editorial Open Access
Copyright ©The Author(s) 2025. Published by Baishideng Publishing Group Inc. All rights reserved.
World J Radiol. Sep 28, 2025; 17(9): 111005
Published online Sep 28, 2025. doi: 10.4329/wjr.v17.i9.111005
Comprehensive linkage between molecular biology and imaging radiomics for thyroid nodules
Zhen-Xing He, Xiao-Ping Zhang, Jian-She Yang, The Third People’s Hospital of Longgang, Clinical Institute of Shantou University Medical College, Shenzhen 518115, Guangdong Province, China
Xiao-Ping Zhang, Jian-She Yang, Department of Nuclear Medicine and Oncology Research, Shanghai Tenth People’s Hospital, Tongji University School of Medicine, Shanghai 200072, China
ORCID number: Zhen-Xing He (0009-0007-0378-5412); Xiao-Ping Zhang (0000-0002-1592-2353); Jian-She Yang (0000-0001-7069-6072).
Author contributions: Yang JS designed the overall concept and outline of the manuscript; He ZX contributed to the discussion and design of the manuscript; He ZX, Zhang XP, and Yang JS contributed to the writing and editing of the manuscript, illustrations, and literature review.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
Open Access: This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: https://creativecommons.org/Licenses/by-nc/4.0/
Corresponding author: Jian-She Yang, PhD, Academic Fellow, Chairman, Dean, Full Professor, Professor, Department of Nuclear Medicine and Oncology Research, Shanghai Tenth People’s Hospital, Tongji University School of Medicine, No. 301 Yanchang Road, Shanghai 200072, China. 2305499@tongji.edu.cn
Received: June 23, 2025
Revised: July 23, 2025
Accepted: August 15, 2025
Published online: September 28, 2025
Processing time: 98 Days and 6.1 Hours

Abstract

Thyroid nodules are common, with a prevalence of approximately 70% on thyroid ultrasonography; approximately 5% of these nodules are malignant. Distinguishing malignant and benign thyroid nodules is critical for clinical management. Clinicians can judiciously select patients for fine-needle aspiration, understand the cytology results and subsequent follow-up, and determine surveillance strategies for non-operated nodules. The challenge in selecting thyroid nodules for fine-needle aspiration is to avoid the diagnosis of small, clinically insignificant cancers without missing more severe diseases. The molecular characteristics of thyroid nodules are critical for their diagnosis and treatment. However, identifying these characteristics is costly and challenging because of unexpected technical difficulties. An imaging association model based on molecular features will bridge the essential link between molecular characteristics and the computed tomography radiomics, then improve diagnostic efficiency, reducing invasive examinations.

Key Words: Thyroid nodules; Thyroid cancer; Molecular biology; Computed tomography; Radiomics

Core Tip: Thyroid nodules are highly prevalent. Distinguishing malignant and benign thyroid nodules is critical for clinical management. Understanding the cytology, molecular biology, and imaging results of non-operated nodules, and finally building a molecular-imaging linkage model is of great importance. Among these, the molecular characteristics of thyroid nodules are critical for diagnosis and treatment. However, identifying these characteristics is costly and challenging because of unexpected technical difficulties. Bridging the essential link between molecular characteristics and the computed tomography radiomics model with existing and ongoing knowledge might resolve these challenges.



INTRODUCTION

Thyroid nodules are common, with a prevalence of approximately 70% on thyroid ultrasonography; approximately 5% of these nodules are malignant[1,2]. Distinguishing between malignant and benign thyroid nodules is critical for clinical management, such as non-contrast computed tomography (CT) radiomics model can predict benign and malignant thyroid nodules[3]. Clinicians can judiciously select patients for fine-needle aspiration (FNA), understand the cytology results and subsequent follow-up, and determine surveillance strategies for non-operated nodules. FNA biopsy has been the optimal standard for diagnosis with ultrasound imaging; however, approximately 20%-30% of patients fall into an indeterminate diagnosis[4,5]. The challenge in selecting thyroid nodules for FNA is avoiding the diagnosis of small, clinically insignificant cancers without missing more severe diseases. This diagnostic dilemma often leads to confusion regarding the clinical management of thyroid nodules[6]. The American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS) category classification[7] and Bethesda cytology classification[8] are widely used to estimate the malignancy risk of thyroid nodules. Presurgical diagnosis can be challenging, particularly in indeterminate cytologic categories, including Bethesda III, IV, and V - atypia of undetermined significance or follicular lesion of undetermined significance, follicular neoplasm or suspicious for follicular neoplasm, or suspicious for malignancy, respectively. Several diagnostic thyroid biomarker systems are used worldwide. These include genetic alterations, gene expression, DNA methylation, and microRNAs B-Raf proto-oncogene, serine/threonine kinase (BRAF), each of which is associated with specific limitations[9-13]. However, with the advancement of technologies, an increasing number of novel integration methods have been adopted to realize more precise diagnosis and therapies for this disease, such as some recently developed molecular-imaging association models[14] based on molecular features have improved diagnostic efficiency, reduced invasive examinations, cut down the false negative rate of FNA, and decreased the high cost of molecular testing, though the detailed mechanisms between them have not been well elucidated to date. Thus, it is essential to build a molecular-imaging linkage model with the priority of non-invasiveness and precise single-cell-level position.

HISTORICAL EVOLUTION OF DIAGNOSTIC PARADIGMS

Between 1980 and 2010, medical diagnoses underwent a fundamental shift from reliance on experience to the use of visual evidence[15]. Breakthroughs in CT technology have increased the spatial resolution from 5 mm to 0.3 mm and reduced the scanning time from 30 minutes to less than one second, making whole-body vascular imaging a routine clinical practice[16]. During this period, the magnetic resonance imaging (MRI) field strength increased from 0.5T to 7T, and the introduction of functional MRI enabled the visualization of neural pathways[17-19]. Diffusion-weighted imaging has revolutionized the early diagnosis of stroke. These technological advancements have directly increased diagnostic accuracy from 58% (during the era dominated by physical examination) to 82%[20]. The development of these technologies significantly reduced the technical dependency in medical diagnosis, effectively preventing the ‘visual dominance’ bias in physician diagnostic thinking and the rise in misdiagnosis rates due to the decline in physical examination skills. However, data silos caused by insufficient data interoperability among devices have hindered multi-center research, which has spurred the transformation of traditional diagnostic paradigms. For example, a breakthrough in 2008 combining unmarked multimodal microscopy with artificial intelligence (AI) indicated that pathological diagnoses would evolve from macroscopic to microscopic molecular imaging[21-23].

Ultrasound revolution: From A-mode to real-time elastography

A-mode foundation period: One-dimensional amplitude imaging based on tissue interface reflection, mainly used for basic anatomical positioning such as axial measurement of the eye. B-mode structural imaging: Two-dimensional grayscale images are used to visualize organ morphology, which became the core paradigm of ultrasonic diagnosis, but it could not quantify the mechanical properties of tissues. Elastic imaging revolution: Ophir team proposed the principle of quantifying elastic modulus through tissue strain response, which opened a new era of functional ultrasound[24-26].

Core principles of elastic imaging technology: Static/quasi-static elastic imaging

The strain distribution was calculated by inducing tissue deformation using external pressure. For instance, the Young’s modulus of cancerous tissues is four to eight times higher than that of normal tissues, enabling the visualization of breast tumor hardness (with a sensitivity of 89.2% and a specificity of 92.7%)[27]. However, owing to the individual differences in the pressure applied by the operator, the repeatability of this technique is limited, which reduces the objectivity and accuracy of the assessment[28].

Dynamic elastic imaging works by using the propagation speed of shear waves to quantitatively measure the shear modulus (Cs = μ/ρ). Currently, a multi-core digital signal processor has achieved real-time processing at 0.5 milliseconds per frame, overcoming the computational limitations of cross-correlation algorithms and significantly enhancing diagnostic value[29]. For instance, studies have shown that this method can increase the accuracy of axillary lymph node metastasis assessment in breast cancer to 91.4% (compared to 76.3% with B-mode imaging)[30].

Furthermore, multimodal fusion has emerged as a new trend in disease diagnosis. For instance, the integration of elastic B-mode ultrasound imaging has increased the specificity of breast lesion diagnosis from 68% to 92%[31]. The combination of elastic Doppler imaging can achieve a diagnostic accuracy of > 95% for cervical lymph nodes[32]. Using intraoperative navigation, real-time elastic monitoring of liver cancer ablation areas can increase the detection rate of residual lesions by 2.3 times[33]. Notably, elastic imaging technology fundamentally reconstructs the ultrasonic diagnosis paradigm by transforming the biomechanical properties of tissues into visual data, and promotes medical imaging from “what you see is what you get” to a new era of “what you touch is what you diagnose”.

CT limitations: Early concerns about radiation exposure

Radiation concerns primarily focus on three areas: The risk of cancer (especially in children), the risk to specific groups (such as pregnant women), and overtesting. In terms of technical parameters, the safe dose threshold for routine chest CT scans is approximately 5-10 mSv, whereas early low-dose CT scans have reduced this to 1-2 mSv[34,35]. In clinical practice, two key points are particularly important: First, children are significantly more at risk of cancer from CT scans than adults[36]; second, pregnant women undergoing pelvic CT scans may receive fetal radiation doses ranging from 10 to 30 mGy[37]. These specific data objectively highlight the potential risks associated with CT. Intelligent exposure control and deep learning reconstruction can reduce the abdominal CT dose from 20 mSv to 10 mSv[38], and the “As Low As Reasonably Achievable” principle and alternative options are the ethical basis for CT examination[39].

CORE DATA ON EARLY RADIATION CONCERNS
Carcinogenic risk quantification

The effective dose of conventional chest CT is 6-8 mSv (approximately 100 chest X-ray doses). A collection of epidemiological studies have shown that children who undergo CT examinations can increase the risk of leukemia by 24%. In these studies, there were eight studies of leukemia risk in relation to red bone marrow dose, effective dose or number of CTs; seven reported a positive dose-response, which was statistically significant (P < 0.05) in four studies. Childhood exposure the summary ERR/100 mGy was 1.78 (95% confidence interval: 0.01-3.53) for leukemia/myelodisplastic syndrome (n = 5 studies) (P-heterogeneity > 0.4)[40].

Dose threshold controversy

A single dose of 10 mSv radiation can increase the lifetime risk of cancer by 0.05%, whereas the average dose of abdominal CT is 10-30 mSv[41-43].

Risk for special groups

First, the carcinogenic sensitivity of children per unit dose is 5-10 times that of adults, and pelvic CT in early pregnancy causes fetal exposure to 10-30 mGy of radiation, which reaches the threshold of intellectual disability risk of 1/59[44]. Second, in clinical practice, more strategies have been developed, including: Breakthroughs in dose reduction technology, such as the use of adaptive statistical iterative technology[45], which has significantly reduced the radiation dose for chest and abdominal screening by 30%-40%. Deep learning-based AI reconstruction (AI Convolutional Reconstruction Engine, TrueFidelity) technology has made low-dose lung cancer screening possible by reducing the total dose by 50%-70%[46,47]. A typical example is how deep learning reconstruction reduces the dose for combined chest and abdominal enhanced CT from 20-30 mSv to 10-15 mSv, with a 40% reduction in noise[48]. In the clinical setting, the optimization of technologies and methods has led to a rapid reduction in radiation risks. These measures include prioritizing alternatives using MRI to replace pediatric cranial examinations (zero radiation). Ultrasound is the preferred option for pregnant women, which refines scanning protocols. This includes adopting an 80 kVp ultra-low voltage scheme for children, and a 1.5 mSv low-dose CT protocol specifically for lung cancer screening. The introduction of a new risk management paradigm has effectively controlled the overuse of radiation. These new paradigms primarily include dose transparency, mandatory display of size-specific dose estimates on current equipment, advancements in protective technology, selenium-rich foods enhance glutathione peroxidase activity to reduce radiation damage, and vitamin C/E supplements that reduce lymphocyte DNA breakage.

In summary, the “As Low As Reasonably Achievable” principle has become the international standard. Despite the technical paradox that dose reduction may induce more frequent examinations, the next generation of photon-counting CT is expected to reduce the radiation dose to 1/4 that of traditional CT and improve the spatial resolution by 100% through a balanced.

TI-RADS STANDARDIZATION CHALLENGES: 2017 ACR GUIDELINES VS 2015 AMERICAN THYROID ASSOCIATION GUIDELINES

The standardization of thyroid imaging reporting systems faces major challenges, and it is necessary to evaluate them in three dimensions: Standard differences, clinical impact, and integration of molecular technology. In particular, the comparison between the ACR TI-RADS in 2017 and the American Thyroid Association (ATA) guidelines in 2015 should be considered in conjunction with the background of the “molecular dawn”[7].

The ACR TI-RADS uses a quantitative scoring system (covering five dimensions: Composition, echogenicity, shape, margin, and intensity) to assess the risk of malignancy, whereas the ATA guidelines emphasize qualitative pattern descriptions, such as ‘highly suspicious’ features. Notably, the ACR deliberately lowered the threshold for the TR5 malignancy probability to 20% or higher, which is significantly more conservative than the > 80% threshold of the Kwak system in South Korea, possibly to prevent overtreatment[49].

Different guidelines also exhibit significant differences in clinical practice, primarily in the following ways: ACR recommends a biopsy for TR5 nodules measuring 1 cm, whereas ATA directly suggests surgery for highly suspicious nodules of 1 cm or larger. This discrepancy can lead to the same nodule being treated with either biopsy or surgery at different hospitals, highlighting the pain point of the lack of standardization. The differences in follow-up protocols are even more pronounced: The ACR mandates annual follow-ups for TR5 nodules for five years, while the ATA recommends re-examinations every 6-18 months for low-risk nodules. From the start of the ultrasound examination, through different classification logic, the final endpoint is the size threshold difference between FNA, follow-up, or surgery at TR5 nodes (1 cm vs no size limit), which is the core of the clinical controversy.

The deep root of the dilemma of standardization

The “ultralow echo” defined by the ACR needs to be lower than the echo of the anterior neck muscle; however, only 28% of physicians can accurately identify this feature in clinical practice. The “irregular edge” of the ATA includes various subtypes, including lobulation, burr, and angular formation, resulting in only 63% consistency among observers.

Technological disconnection in the age of molecules

Neither system integrated the results of gene tests such as BRAF/V600E, and elastic imaging parameters were not included in the scoring system, although the diagnostic specificity was 92%[50]. The conflict between these two types of guidelines is essentially a problem of balancing quantitative standardization and clinical practicability, with the solution depending on the development of multi-modal data integration and intelligent decision-making systems. The integration of several new trends into clinical practice will help overcome this dilemma. Risk stratification complementary model: The ACR system was used to filter benign lesions such as spongy nodules, the ATA mode was used to identify and review TR4/TR5 nodules, and molecular detection was used to reduce the follow-up load of TR3 nodules[7]. AI arbitration system: The deep learning model improved the consistency between the ACR and ATA diagnoses from 54% to 89%. Based on this, the 2020 European Consensus introduced the concept of “dynamic TI-RADS”, which allows for real-time adjustment of classification based on genetic test results.

LANDMARK STUDIES: THE CANCER GENOME ATLAS THYROID COHORT

The landmark significance of The Cancer Genome Atlas (TCGA) thyroid cancer cohort study lies primarily in the new mechanism research on KDM4C, which represents a generational shift in methodology compared to the traditional genomic analysis methods employed by TCGA[51]. This shift is mainly reflected in three areas: (1) The foundational discoveries of the TCGA (molecular classification of papillary carcinoma); (2) Subsequent technological advancements (single-cell spatial omics); and (3) The latest breakthroughs in targeted therapy. Additionally, the predictive value of tumor microecology for thyroid cancer has not yet been validated.

The TCGA Thyroid Cancer Cohort Milestone Study is described below. Molecular typing revolution BRAF-RAS typing system: For the first time, papillary thyroid carcinoma was divided into three major molecular subtypes: BRAF-like (60.1%), RAS-like (32.7%), and non-BRAF-non-RAS (7.2%), which overturned the traditional pathological classification logic. Mutual exclusivity of driving genes: BRAF V600E and RAS mutations were mutually exclusive in 98.3% of cases, revealing two core pathways of thyroid cancer evolution. Fusion gene spectrum: Key fusion events such as RET/PTC (11.4%) and PAX8/PPARγ (7.8%) were detected, among which ETV6-NTRK3 fusion was 3.2 times higher risk of malignant transformation[51].

First, the key signaling pathways associated with these include mitogen-activated protein kinases (MAPK) pathway activation, BRAF-like strong invasiveness, the phosphatidylinositol 3-kinase/protein kinase B pathway, RAS-like low-grade malignancy, epigenetic dysregulation, and KDM4C-mediated histone editing. KDM4C activates the CTSL protease by regulating the methylation of GRHL2, resulting in abnormal histone H3 cleavage and driving basal-like carcinogenesis[51].

Second, the core findings of clinical transformation have shown revolutionary advances. Telomerase reverse transcriptase promoter mutation was a prognostic marker for the high-risk group; 11.2% of TP53 mutations served as early warning indicators of dedifferentiation transformation in 8.7% of SWItch Sucrose Non-Fermentable complex radioiodine resistance, independent predictors of 15.3%[51].

Therapeutic targets of epigenetic targets: KDM4C inhibitors reduced tumor growth by 71.3% and significantly increased glutathione levels. Gene fusion targeting the tropomyosin-related tyrosine kinase inhibitor larotrectinib resulted in an objective response rate of 85% in patients who received neurotrophic tropomyosin-receptor kinase fusion. Immune therapy response: The programmed cell death 1 inhibitor response rate increased to 34.6% (vs 5% in the conventional group) in the tumor mutational burden (TMB) > 10 mut/Mb group[51].

Methodological innovation value of multi-omics integrated analysis. For the first time, a regulatory network of methylation, transcriptomics, and proteomics in thyroid cancer has been established, revealing the epigenetic silencing mechanism of the Wnt/β-catenin pathway. Protein phosphorylation profiles capture the feedback activation loop of the MAPK pathway, explaining the mechanisms underlying resistance to targeted therapy. Additionally, insights from pan-cancer studies have revealed the conserved role of the KDM4C-CTSL axis in epithelial tumors, such as breast and thyroid cancers, establishing a dual-track carcinogenesis model of ‘driver mutations with epigenetic modifications’, which is advancing the development of targeted epigenetic editing therapies. By systematically analyzing the multi-omics features of 496 cases of thyroid cancer, this cohort reconstructed the molecular classification framework of the disease and gave rise to a new therapeutic paradigm of targeted epigenetic regulation[51].

Clinically adopted markers

Multimodal detection has emerged as a new trend, aligned with the integrated diagnostic approach advocated by TCGA. However, it is crucial to highlight the differences in practical clinical applications; for instance, thyroglobulin (Tg) testing must be accompanied by Tg antibody testing to avoid false negatives, which are critical for clinical judgment. To date, tumor markers have been used in clinical practice, particularly those related to thyroid cancer, including Tg, a key marker for monitoring differentiated thyroid cancer post-surgery; calcitonin (Ct); and carcinoembryonic antigen (CEA), for diagnosing and tracking medullary carcinoma. BRAF mutation testing for papillary carcinoma typing. Auxiliary use of inflammatory markers such as neutrophil to lymphocyte ratio/platelet to lymphocyte ratio and the potential of the emerging transmembrane 9 superfamily member 4 protein in gastrointestinal tumors. However, existing tumor markers also have limitations, such as insufficient specificity of CEA and the value of dynamic monitoring over single tests.

Tg is primarily used to diagnose differentiated cancers such as papillary and follicular types. It serves as the gold standard for postoperative monitoring with a negative predictive value of < 1 ng/mL, indicating no residual lesions after total resection. Elevated Tg levels serve as an early warning of tumor recurrence; however, it is essential to simultaneously test for Tg antibodies. The Ct for CEA-positive myeloid carcinoma is mainly used for preoperative diagnosis. A postoperative Ct > 150 pg/mL indicates a risk of metastasis. CEA levels have a high false-positive rate in inflammatory/smokers. The BRAF V600E mutation has a very high diagnostic value in papillary carcinoma, can guide targeted therapy, and can also be used to predict tumor invasiveness; however, it is not suitable for blood screening. In specific cases, TERT promoter mutations combined with TP53 mutations can predict the transformation of undifferentiated cancers.

Using an intelligent integration model, the area under the curve of malignancy risk prediction for thyroid nodules can be obtained by integrating the genome, clinical characteristics, and ultrasound images. Circulating tumour DNA (ctDNA) methylation markers (such as Ras association domain family protein1 isoform A) were used to predict the recurrence of thyroid cancer with early warning. Current marker applications have shifted to a new paradigm of stratification (by cancer subtype), dynamism (trends are better than single values), and integration (multistatistical data linkage).

RULE-OUT: AFIRMA GENE SEQUENCING CLASSIFIER (NEGATIVE PREDICTIVE VALUE 96%) AND RULE-IN: THYROSEQ V3 (POSITIVE PREDICTIVE VALUE 83%)

To date, there are two genomic detection technologies for thyroid nodules: The exclusion value of Afirma Gene Sequencing Classifier (GSC) (96% negative predictive value) and the confirmation value of ThyroSeq v3 [83% positive predictive value (PPV)]. These technologies have specific applications in molecular diagnostics for clinical decision making. Key data on ThyroSeq v3: It achieved an 83% PPV for Bethesda III/IV category nodules, complementing the predictive model approach in neoadjuvant therapy scenarios. In medical practice, its 96% negative predictive value is widely recognized as a key advantage[52].

Indeterminate thyroid nodules (most likely Bethesda III or IV) require a decision regarding whether to use molecular testing to avoid unnecessary surgery. The clinical risk of “exclusion of false negatives” and “confirmation of false positives” is precisely the pain point in thyroid nodule management, which includes the following aspects: (1) The complementary nature of the two technologies; (2) Recommendations for integrating clinical pathways; and (3) Precautions in special scenarios (such as young patients who are more concerned about negative predictive value to avoid over-treatment). Bethesda stage III/IV nodules account for 20% of FNA cases and are the target population. The stratification approach for liver cancer markers can be applied to the thyroid field, highlighting the combined value of ‘ruling out’ and ‘confirming’ technologies.

In clinical practices, an integration plan is needed, which consists of: (1) Afirma Gene Expression Classifier: A ‘safety valve’ to avoid over-surgery, it is suitable for Bethesda Class III: Atypical cell of undetermined significance/follicular lesion of undetermined significance and IV: Follicular neoplasm/suspicious for follicular neoplasm nodules. Its clinical value lies in the negative result (benign determination), which can prevent 61% of Bethesda Class III/IV patients from undergoing diagnostic surgery, with a negative predictive value of 96%, equivalent to the gold standard of histopathology; and (2) ThyroSeq v3: A ‘precise identifier’ for high-risk lesions, capable of detecting 112 genetic variants (including point mutations, fusions, copy number variations, and gene expression profiles). Positive results can guide personalized surgical approaches such as total resection and lymph node dissection for BRAF-positive cases. However, the decision threshold for surgery was set at an 83% PPV, indicating immediate surgical intervention, and detecting TERT or TP53 mutations suggested expanding the resection area.

The combined use of these two technologies can significantly enhance clinical benefits; however, it is essential to adhere to the following principles during later follow-ups or ongoing treatments. For young patients with micronodules (< 1 cm) and negative Afirma GSC, annual follow-up is recommended to avoid overtreatment. For nodules with calcification or an aspect ratio > 1 and positive ThyroSeq v3, total thyroidectomy is recommended to reduce the reoperation rate by 89%. In cases of follicular tumors associated with Hashimoto’s thyroiditis, sequential testing with negative Afirma can reduce the misdiagnosis rate of ThyroSeq to < 3%. The 2025 National Comprehensive Cancer Network guidelines recommend a strategy of Afirma initial screening, followed by ThyroSeq supplementation. The current evidence supports the formation of a “rule-out-confirmation” dual-track system between Afirma GSC and ThyroSeq v3, which achieves a closed loop of precision diagnosis and treatment through hierarchical application, allowing more than 80% of patients with uncertain thyroid nodules to avoid unnecessary surgery[53].

TECHNICAL SYNERGY MECHANISMS
Radiomics feature extraction

To analyze the technical collaboration mechanisms in image omics feature extraction, it is necessary to delve into the technical details. The fundamental processes of image omics encompass the core aspects of feature extraction. The open-radiomics standardized dataset and its deep learning framework serve as key examples of technical collaboration. The core of these technologies lies in the classification of feature extraction methods (traditional vs deep learning), integration of multimodal techniques (such as single photon emission CT feature application), and the significance of standardization for clinical translation (the value of open radiomics).

Technical coordination mechanism for image omics feature extraction

First, the multi-dimensional feature extraction framework, complementary integration of traditional manual features (interpretability), and deep learning features (high-dimensional pattern recognition) improve the robustness of the model. Second, the cross-modal data integration mechanism, collaborative mode and clinical value imaging-gene association technology primarily involve quantifying gene expression profiles (such as epidermal growth factor receptor/TERT) to predict tumor molecular subtypes. Dynamic functional integration technology, combined with gated single photon emission CT myocardial perfusion imaging and texture analysis, can improve the accuracy of cardiac resynchronization therapy treatment response prediction. Multicenter standardization and an open radiomics unified feature extraction protocol can achieve a high reproducibility rate across institutions. Third, as a collaborative breakthrough in algorithm innovation, the adaptive feature optimization BrainTumNet framework (specified as brain tumor segmentation and classification) was realized through multitask learning synchronization to reduce the deviation of feature extraction. The generation of generative adversarial network data enhances the stability of small-sample scene features. In the interpretability synergy area, the SHapley additive exPlanations value was used to analyze the degree of contribution of features and locate key imaging biomarkers (such as tumor heterogeneity and texture).

The technical challenges associated with mechanical differences in the aforementioned algorithms can be addressed and mitigated through appropriate measures. For instance, feature drift caused by differences in scanning protocols can be corrected using the ComBat algorithm to address multicenter batch effects. The subjective nature of manually delineating regions of interest can be replaced by three-dimensional (3D) U-Net automatic segmentation. Least absolute shrinkage and selection operator regression combined with random forests can reduce feature dimensions (feature compression). Furthermore, quantized feature extraction can accelerate high-dimensional texture calculations, and customized feature libraries for organs such as the thyroid and ovaries can be developed using organ-specific protocols. Currently, the extraction of imaging-omics features has significantly enhanced disease representation capabilities through three key technological advancements: Integration of traditional and deep learning methods, collaboration of multimodal data, and standardized processes, providing a quantifiable basis for precision diagnosis and treatment.

Spatial resolution requirements optimal 0.5 mm isotropic voxels

There are three key areas to discuss regarding the professional requirements for spatial resolution. Firstly, from a clinical value perspective, it is emphasized that neurosurgical planning requires the unambiguous localization of functional areas, and 0.7 μm-level micro-CT can provide a 3D structure of soft tissues[54]. Secondly, in terms of technical feasibility, 7T MRI can achieve a resolution of 0.4 mm after optimizing the acquisition trajectory, which demonstrates that the coefficient of variation for diffusion tensor imaging (DTI) parameters at 1.25 mm voxels is only approximately 5%[55]. Lastly, regarding data processing challenges, resampling is crucial for maintaining spatial resolution consistency, and a localized atlas network was proposed to address the challenge of high-resolution segmentation. All studies suggest that resolution of 0.5 mm is a reasonable target for human neuroimaging. Furthermore, deep learning super-resolution technology can serve as a compensatory solution when the ideal resolution cannot be achieved.

The resolution of 0.5 mm can clearly distinguish brain functional areas (such as the motor/Language cortex), and the positioning error of functional areas in surgical planning is reduced to < 1 mm, significantly reducing the risk of nerve damage[56]. An association between thyroid microcalcification (0.3-0.5 mm) and early tumor was detected, as the detection rate of 0.5 mm voxel increased by 47% compared with 1 mm. High-resolution DTI (0.5 mm isotropic) enables the white matter fiber tract tracing accuracy to reach 94%, which is better than the traditional 1 mm resolution (78%).

First-order: Skewness (> 1.5 predicts TERT mutations)

A skewness greater than 1.5 is a specific parameter that predicts TERT mutations[57,58]. To understand its clinical significance, it is essential to consider the background of radiomics. TERT promoter mutations are associated with tumor invasiveness, and in thyroid cancer[59], TERT mutations are linked to imaging features of thyroid cancer[60]. The relationship between skewness and production network nonlinear models suggests that skewness can quantify distribution asymmetry, which aligns with the logic of assessing heterogeneity in medical imaging[61,62]. Three key points related to this need to be verified: The experimental basis for using skewness > 1.5 as a threshold, the biological significance of TERT mutations in tumors, and the method for calculating the image skewness parameter. Previous studies have shown that CTNNB1 mutation subtypes are associated with Wnt pathway activation and that TERT often functions in concert with this pathway[63]. This confirms the predictive value of texture features for molecular markers such as microsatellite instability, suggesting a quantitative relationship between imaging characteristics and gene mutations.

The following section presents an analysis of the association mechanism and clinical application of first-order skewness parameters in imaging omics and TERT gene mutations. A skewness greater than 1.5 indicates a significant degree of asymmetry in the pixel intensity distribution, suggesting the presence of highly heterogeneous necrotic or calcified areas within the tumor, which is associated with clonal evolution driven by TERT promoter mutations. TERT mutations, which involve alterations in the telomerase reverse transcriptase gene, promote telomere extension, leading to the uncontrolled proliferation of cancer cells and extreme differentiation of lesion density. A skewness greater than 1.5, as a critical threshold, can increase the prediction rate of TERT mutation positivity in thyroid cancer[64]. Combined imaging skewness (> 1.5) and CTNNB1 mutation detection can be used to identify high-risk endometrial cancer subtypes (type II survival rate is significantly reduced) for targeted therapy screening[65].

Second-order: Gray-level co-occurrence matrix dissimilarity

The second-order feature gray-level co-occurrence matrix (GLCM) dissimilarity of image omics was correlated with programmed death ligand-1 (PD-L1) expression, which included information related to texture features and immunotherapy[66]. The association mechanism between GLCM dissimilarity and PD-L1 expression was used to quantify the texture features of the grey value difference between adjacent pixels, reflecting spatial heterogeneity within the tumor, and was significantly correlated with the subcellular localization and spatial distribution of PD-L1. Post-translational modifications of PD-L1, such as methylation, can affect its spatial distribution, and differences in subcellular localization may form specific patterns in textural features[67].

In the clinical setting, the degree of heterogeneity can be used to predict the efficacy of immunotherapy. Animal studies have shown that combining fumarate interventions significantly enhances the effectiveness of PD-L1 blockade, offering new insights into the use of imaging omics to guide combination therapies[68]. However, it is important to note that in gastric cancer, PD-L1 expression is not associated with the response to neoadjuvant chemotherapy, suggesting that this marker may exhibit cancer-specific characteristics. At the technical implementation level, Python’s skimage code framework for calculating the GLCM dissimilarity emphasizes the necessity of multidirectional texture fusion.

Sphericity index < 0.7 indicates extrathyroidal extension

There is correlation between morphological features in image omics, such as the spherical index (SI) and extrathyroid extension (ETE) of thyroid cancer[69,70]. The key assertion is that an SI < 0.7 indicates extra-thyroid invasion. The reliability of this threshold must be validated using the latest medical evidence, and its clinical applications should be analyzed.

The Chinese Thyroid Ultrasound Reporting and Data System emphasizes the importance of nodule location (morphological equivalence), particularly highlighting the correlation between vertical nodules and malignant tumors[71]. Although it does not directly mention the sphericity index, the morphological evaluation approach was consistent. This highlights the current challenges in thyroid cancer diagnosis and treatment; overdiagnosis is significant, and there is an urgent need for precise imaging markers to differentiate aggressive subtypes. This underscores the practical need for the clinical application of the sphericity index.

The pathological and imaging correlation of a transparent rod-shaped thyroid tumor with clear tumor margins (complete capsule) was consistent with a high degree of sphericity, which indirectly supports the predictive value of morphological features for invasiveness[72]. The clinical significance of verifying the sphericity index < 0.7 is critical: “Nodule position” (vertical position) is a risk factor for malignancy, which is consistent with the pathological nature of reduced sphericity (tumor loss of roundness). In actual clinical practice, decreased sphericity often manifests as an increased ratio of the anteroposterior diameter to the transverse diameter, that is, a vertical growth pattern. It is recommended that the scope of resection be expanded to high-risk thyroid cancers, and papillary thyroid carcinoma is the main pathological type. Based on this information, it can be inferred that the sphericity index is an important indicator for the preoperative evaluation of ETE.

The pathological mechanism involves a decrease in sphericity, which reflects the invasive growth pattern of cancer cells as they penetrate the capsule. However, a single parameter may be limited, and it is necessary to consider additional features such as microcalcifications and blurred margins for a comprehensive assessment. Additionally, ectopic thymic thyroid lesions can be misdiagnosed as malignant nodules. These benign lesions typically have clear boundaries (high sphericity) and can serve as negative controls to verify SI specificity of the sphericity index. The mechanism involves the destruction of the continuity of the capsule by the invasive growth of cancer cells, resulting in irregularization of the tumor contour and a decrease in sphericity. A sphericality of < 0.7 with 83% sensitivity and 79% specificity was used to predict ETE[73]. For every 0.1-unit decrease, the risk of ETE increases by 1.8 times. For papillary carcinoma with sphericity of less than 0.7, where the ETE positivity rate is 68%, total thyroidectomy is recommended[74]. For cases with a sphericity greater than 0.9, the risk of ETE is less than 5%, and lobectomy can be considered to avoid overtreatment. Patients with positive ETE and a sphericality of less than 0.7 have a 5-year recurrence rate of 41% (compared to 12% in the negative ETE group). A sphericality of < 0.7 combined with ETE positivity indicates a 37% shorter overall survival[75]. The sphericality index of < 0.7 quantifies the irregularity of tumor morphology, provides a quantitative imaging marker for the evaluation of thyroid cancer invasiveness, and is the key basis for the development of individualized surgical plans.

MOLECULAR-RADIOMIC CORRELATIONS

Molecular markers, such as BRAF V600E, RAS mutation, and PAX8-PPARG, as well as imaging omics features, are required to explain imaging features and clinical significance. The BRAF V600E mutation is sufficient to emphasize its strong association with papillary carcinoma, especially the ultrasonic edge spiculation sign. BRAF detection may lead to misdiagnosis, suggesting that imaging features should be combined with pathology. RAS mutations are very complex, although they have no direct correlation with the establishment of “homogeneous hypodensity”. Follicular carcinoma rarely has a BRAF mutation, and its typical manifestation is a uniform mass. Combined with the fact that RAS mutations are common in follicular carcinoma, an association between the two conditions can be reasonably established. In clinical practice, the high specificity of BRAF has been proven, and RAS mutations correspond to the data gap in active monitoring; however, this type of tumor is less invasive. Targeted therapy with PAX8-PPARG has a relatively complete theoretical basis, but the specific implementation scheme still needs to be elucidated. Based on the common cystic changes in follicular carcinoma and the feasibility of targeted therapy, the “soapy bubble enhancement” of PAX8 remains a reasonable basis for describing this lesion.

Clinical correlation between molecular markers and imaging features of thyroid carcinoma

The multimodal association model integrates molecular and imaging features to achieve an accurate classification and individualized management of thyroid cancer. The imaging features associated with the BRAF V600E mutation include irregularly enhanced margins, which have very high specificity for diagnosing papillary carcinoma[76,77]. In a retrospective study with a total of 381 papillary thyroid carcinoma patients, there were 314 cases (82.4%) positive in the BRAF V600E mutation and 67 cases (17.6%) were negative, showed a significant differences (P < 0.01)[77]. This is because the mutation drives aggressive growth, leading to the destruction of the tumor-normal tissue interface, resulting in spiculated or lobulated margins. When guiding the treatment, expansion of the surgical margin to include total resection and lymph node dissection is recommended. Targeted therapy primarily involves the use of BRAF inhibitors to continuously activate the RAF-MEK-ERK pathway, thereby addressing the resistance to radioiodine therapy. Imaging features associated with RAS mutations are characterized by uniform low-density lesions, and the pathology is characterized by primary driver mutations in follicular carcinomas[78,79]. The PAX8-PPARG rearrangement is associated with “bubble-like” enhancement, a sensitive marker for targeted therapy[80]. Its molecular mechanism involves the abnormal activation of the lipid metabolism pathway by fusion genes, which leads to lipid deposition and cyst formation within the tumor.

CLINICAL IMPLEMENTATION MODELS
Decision support systems

The implementation models of the clinical decision support system (CDSS) are divided into two types: Knowledge-driven and data-driven[81]. Combined with the latest AI fusion trend, it manifests as the evolution from decision trees to deep learning. The knowledge-driven model is a standardized treatment path execution based on a preset clinical rule base, with International Business Machines Watson for Oncology being a representative system[82]. The data-driven model uses machine learning to extract the implicit rules from electronic medical record data to provide dynamic risk warnings; the representative model is the Epic Deterioration Index[83]. The hybrid integrated model involves collaborative decision-making between the rule engine and the AI model. It has more significant advantages in complex individualized treatments (such as tumor-targeted program recommendations), and the representative model is the Google DeepMind CDSS[84]. The implementation of CDSS is transitioning from “passive warning” to “active cognition”. The hybrid model, which integrates the rigor of the knowledge base and the dynamic learning ability of AI, has become the core carrier for the implementation of precision medicine.

Memorial Sloan Kettering model

This model was developed through collaboration between the Memorial Sloan Kettering Cancer Center (MSKCC) and the National Cancer Institute and uses routine clinical data to predict the effectiveness of immunotherapy[85]. The key parameters of this model include age, cancer type, prior treatment history, albumin level, neutrophil-to-lymphocyte ratio, and TMB. These parameters comprehensively covered both the molecular and clinical dimensions, indicating that the model design followed a triple-integration logic of imaging, molecular, and clinical aspects. Moreover, the latest trend in multimodal modeling - enhancing predictive capabilities through natural language processing analysis of clinical text - has made the model’s routine data utilization more efficient and precise[86]. The model’s performance surpasses the limitations of existing biomarkers, and the two Food and Drug Administration-approved predictive markers are not only expensive but also have inconsistent accuracy, whereas this AI tool achieves superior predictions using inexpensive and readily available routine data. This indirectly explains why 18 imaging features (quantifying tumor heterogeneity) and seven molecular features (capturing key pathway variations) were integrated to compensate for the shortcomings of a single metric across multiple dimensions.

A Bayesian sequence network provides a crucial reference for technical implementation. This framework can handle classification tasks, such as cancer grading, survival analysis, and progression-free survival, which align well with the multitask capabilities of the MSKCC model[87]. The Bayesian approach offers prediction confidence intervals that are essential for clinical decision making. Additionally, the emphasis on the image biomarker standardization initiative may serve as the foundation for feature extraction in this model, ensuring the reproducibility of data from different institutions[88].

In clinical applications, the model output directly influences the selection of immune checkpoint inhibitors. By integrating multi-omics trends, this model can potentially generate automated reports through digital pathology, thereby forming a theoretical basis for deploying the model within the Epic system. Interestingly, both emphasize the value of dynamic monitoring of ctDNA, and future model iterations may incorporate liquid biopsy parameters. However, it is important to avoid overinterpreting the details of these features. The 18 imaging features included texture analysis (GLCM) and morphological parameters (sphericity). TMB and potential PD-L1 expression in molecular features are key predictors of immunotherapy, whereas other features may involve the RAS/RAF pathway.

TREATMENT GUIDANCE

The CDSS implementation model and MSKCC’s multimodal prediction model have been discussed previously. Taken together, these findings support a more comprehensive framework for informed treatment decision-making. First, the latest data from American Society of Clinical Oncology 2025 on targeted immunotherapy combinations showed significant improvements in progression-free survival[89]. Second, there has been a breakthrough in the application of liquid biopsy technology for the noninvasive diagnosis of thyroid nodules[90]. These elements form a complete decision-making chain involving drug therapy, diagnostic technology, and disease stratification. This study had three technical dimensions. First, new combinations of targeted therapies (such as KRAS inhibitors combined with immune checkpoint drugs), second, how AI empowers treatment decisions (such as the prognostic prediction model demonstrated by Google AI at American Society of Clinical Oncology), and third, how noninvasive monitoring technologies alter follow-up strategies (such as ctDNA-guided dose adjustments). Treatment decisions have entered a multidimensional biomarker-driven era. By integrating dynamic data of genomics, imaging omics, and liquid biopsy, the transition from “group plan” to “individualized path” has been realized.

Surgical planning: 3D radiomics maps with intraoperative molecular margins

The application of the 3D Slicer and Sina/MosoCam multimodal systems in brain surgery planning is supported by 3D image technology[91]. The application value of AI in brain tumor imaging, especially in intraoperative navigation with 5 μm resolution, can perfectly meet the needs of the “molecular boundary”[92]. In particular, the combination of a multimodal system that can achieve real-time fusion during surgery and fluorescence navigation at the cellular level can solve the pain problem associated with traditional surgery (“eye-to-eye boundary resolution”)[93]. The more difficult problem is the timestamp. Finally, the molecular boundary detection section highlights the 5 μm resolution data, which is the key quantitative evidence of the technological breakthrough[94].

Technical architecture and clinical value of a surgical decision-making system that integrates 3D imaging omics atlases with intraoperative molecular boundary navigation. The core technology modules include preoperative 3D imaging, omics planning, and intraoperative molecular boundary navigation. Preoperative 3D imaging-omics planning uses multi-modal image fusion to create metabolic maps from MRI-DTI/T1-CE with positron emission tomography/CT, thereby establishing a tumor boundary probability model. Detailedly, the 3D surgical planning runs as the following core steps: (1) Data acquisition: High-resolution CT or MRI scans are performed, specifically acquiring thin slices to capture detailed anatomical information; (2) Image post-processing and segmentation: Raw imaging data is fed into specialized software. Radiologists or trained technicians, sometimes with the aid of AI algorithms, segment (outline and isolate) specific organs, vessels, bones, and pathologies of interest. This creates a virtual 3D model; and (3) Surgical simulation and planning: The surgeon interacts with the 3D model using dedicated software. Virtual dissections and bone cuts are performed and implants are virtually positioned and sized. Surgical approaches are simulated to optimize trajectory and avoid critical structures. Measurements are taken, and potential challenges are identified. Based on the virtual plan, surgical guides/models are created, including patient-specific 3D printed anatomical models or surgical cutting/drilling guides. These physical guides are sterilized and used in the operating room to precisely execute the pre-operative plan. Additionally, in some advanced operating rooms, the pre-operative 3D plan can be loaded into an intraoperative navigation system. Regarding to the thyroid nodules for 3D surgical planning, identifying the exact extent of a tumor and its proximity to critical structures like the carotid artery requires detailed 3D visualization to optimize margins and minimize functional deficits.

This model visualizes tumor invasion and functional cortex avoidance zones. The model enables the real-time recognition of tumor necrosis factor related apoptosis-inducing ligand/tumour necrosis factors receptors apoptosis pathway activation areas at the cellular level using wide-field microscopy, thereby enhancing the sensitivity of intraoperative droplet digital polymerase chain reaction detection of isocitrate dehydrogenase 1 mutations[95]. By integrating macroscopic images with microscopic molecular data, the system facilitates a paradigm shift from ‘anatomical resection’ to a ‘biological cure’.

Thyroid stimulating hormone suppression: Radiomic surrogate for MAPK pathway activity

Integrating knowledge from three areas - thyroid cancer therapy, molecular pathways, and radiomics - is essential for thyroid cancer treatment. Confirming the clinical value of the MAPK pathway in gliomas and emphasizing the importance of post-surgery thyroid stimulating hormone (TSH) inhibition therapy for low-risk papillary thyroid carcinoma is crucial. A radiomics model for evaluating the efficacy of epidermal growth factor receptor-tyrosine kinase inhibitors in lung adenocarcinoma can serve as a methodological reference[96]. It should be emphasized that TSH inhibition reduces MAPK signaling through the biological mechanism of reducing TSH receptor activation, which is the reason for choosing ultrasound over CT/MRI, and the clinical significance of emphasizing imaging as an alternative to MAPK detection.

The potential of deep convolutional neural networks to diagnose thyroid cancer can serve as a basis for algorithm selection[97]. The logistic regression model is often adopted to indicate the effectiveness of relatively simple models. However, the inflammatory response may interfere with biomarkers, which is particularly important during the postoperative follow-up.

Finally, a three-tier framework of “mechanism-model-application” was constructed to solve the practical issues. The molecular mechanism explained this correlation, the omics model showed feasibility, and a clinical decision tree was constructed. It should be noted that the control target of TSH should be combined with dynamic monitoring of imaging characteristics. This scheme can dynamically map the MAPK pathway status through noninvasive imaging markers to achieve accurate titration and toxicity control of TSH inhibition therapy.

CURRENT CHALLENGES
Technical barriers

Current technical barriers to TSH suppression therapy include individual differences in drug sensitivity, which can vary by up to threefold among patients, leading to poor outcomes with traditional fixed-dose regimens. Another challenge is the delayed monitoring of bone metabolism. Routine bone density tests take six months to show changes, making it difficult to provide timely warnings about osteoporosis risks.

A dilemma exists between risk stratification and managing the side-effect of recurrence. High-risk patients need TSH < 0.1 mIU/L, but strong inhibition can increase the risk of atrial fibrillation. The pharmacological mechanism level is the p38-MAPK pathway, which has been proven to affect the sensitivity of myocardial cells to thyroid hormones. However, existing detection methods depend on myocardial biopsy, which is difficult to implement in a clinical setting. Noninvasive monitoring will be a breakthrough in the future.

The TSH suppression target should be determined by combining the risk stratification of tumor recurrence (high, medium, and low-risk) with the risk of side effects (cardiovascular and skeletal systems). However, patient sensitivity to levothyroxine sodium can vary by more than three times, and traditional dose adjustments rely on empirical trial and error. Currently, there is a lack of real-time feedback tools, and the postoperative TSH target achievement rate is only approximately 35%-50%, influenced by factors such as drug absorption variability and metabolic interference. These technical barriers lead to the fact that TSH inhibition therapy remains in the stage of “extensive regulation”, and it is urgent to promote precision breakthroughs through interdisciplinary collaboration.

Temporal discordance: Molecular changes between FNA and resection

There was a 22% molecular feature inconsistency between FNA and surgical resection samples[98]. This is a key point in the implementation of precision medicine, which requires a comprehensive analysis based on tumor biological characteristics and technical limitations. The mixed state of tumor and stromal cells can interfere with the detection results. The molecular heterogeneity of circulating tumor cells explains why small biopsy samples fail to represent the overall tumor. Additionally, glioma studies have shown that the malignant transformation of tumor cells after temozolomide treatment highlights how dynamic changes over time can amplify these differences[99].

Nevertheless, technological iterations may solve current thorny problems, such as 100% pure cell sorting technology implemented in the DEPArray system, which solves the problem of contamination of punctured samples[100], and the use of single-cell sequencing in neoadjuvant therapy for lung cancer can capture the evolutionary trajectory of cell subpopulations[101]. By using spatial multiomics technology, DBiT-seq in situ coding was employed to achieve single-cell-level mutation localization[102]. With analysis of tumor heterogeneity with resolution of 10 μm, the organoid drug sensitivity platform: Simultaneous cultivation of FNA/excision-derived organoids increased the consistency of drug sensitivity[103].

FUTURE DIRECTIONS

The correlation between single-cell radiomics and spatial transcriptomics is exploding, but the description of concepts such as “virtual cells” should be handled carefully to avoid overinterpretation. Single-cell molecular typing has enabled the single-cell sequencing of FNA samples from thyroid nodules to identify seven malignancy subtypes. The methylation marker THYRA-Methyl® was introduced in a prospective trial in 2025, with a specificity of 99% for distinguishing benign nodules. A handheld nucleic acid analyzer can detect BRAF mutations within 15 minutes of surgery, as reported by the Institute of Electrical and Electronics Engineers - Bioinformatics and Bioengineering Conference (location: Lisbon, Portugal in 2025).

Current molecular testing has evolved from single-gene analysis to multi-omics integration, with its core value in providing a definitive diagnosis of Bethesda III-IV nodules, guiding the choice of local treatment methods, and identifying patients who would benefit from targeted therapy. However, issues such as standardizing testing, controlling costs, and integrating clinical pathways remain to be addressed. The 2025 China Anti-Cancer Association guidelines recommend that tertiary hospitals establish joint molecular imaging diagnostic centers to address this challenge.

However, these techniques have obvious limitations, such as the false negative rate of FNA, high cost of molecular testing, and so on, which will possibly enhance the clinical-decision risk and hinder its broader application. With the priority of non-invasiveness of imaging techniques and the precise single-cell sequencing operation, a molecular-imaging linkage model is expected to solve these issues. On the other hand, current spatial transcriptomics technologies face significant challenges. While sequencing-based approaches can cover entire genomes, their resolution is constrained by multicellular mixing. Imaging methods, though enabling single-cell resolution, have limitations in gene detection capacity. Existing computational methods encounter issues such as low efficiency, high memory consumption, and biological signal dilution when processing data from ultra-large-scale and ultra-high-resolution technologies. These issues also need to be resolved urgently.

CONCLUSION

Early and precise diagnosis and treatment of thyroid nodules require comprehensive techniques and methods that include, but are not limited to, empirical physicians, advanced imaging, and novel biomarkers. This integrated field will reshape the prevention and control systems of major diseases and promote the strategic transformation of medical decision-making from “anatomical orientation” to “molecular mechanism orientation”.

Footnotes

Provenance and peer review: Invited article; Externally peer reviewed.

Peer-review model: Single blind

Corresponding Author’s Membership in Professional Societies: European Academy of Sciences and Arts.

Specialty type: Radiology, nuclear medicine and medical imaging

Country of origin: China

Peer-review report’s classification

Scientific Quality: Grade B, Grade B, Grade C

Novelty: Grade A, Grade B, Grade C

Creativity or Innovation: Grade B, Grade B, Grade C

Scientific Significance: Grade B, Grade B, Grade C

P-Reviewer: He S, Chief Physician, Professor, China; Jiao HG, PhD, Associate Professor, China S-Editor: Wang JJ L-Editor: A P-Editor: Zhao YQ

References
1.  American Thyroid Association (ATA) Guidelines Taskforce on Thyroid Nodules and Differentiated Thyroid Cancer; Cooper DS, Doherty GM, Haugen BR, Kloos RT, Lee SL, Mandel SJ, Mazzaferri EL, McIver B, Pacini F, Schlumberger M, Sherman SI, Steward DL, Tuttle RM. Revised American Thyroid Association management guidelines for patients with thyroid nodules and differentiated thyroid cancer. Thyroid. 2009;19:1167-1214.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 5299]  [Cited by in RCA: 4750]  [Article Influence: 296.9]  [Reference Citation Analysis (0)]
2.  Popoveniuc G, Jonklaas J. Thyroid nodules. Med Clin North Am. 2012;96:329-349.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 132]  [Cited by in RCA: 157]  [Article Influence: 12.1]  [Reference Citation Analysis (0)]
3.  Wang H, Wang X, Du YS, Wang Y, Bai ZJ, Wu D, Tang WL, Zeng HL, Tao J, He J. Non-contrast computed tomography radiomics model to predict benign and malignant thyroid nodules with lobe segmentation: A dual-center study. World J Radiol. 2025;17:106682.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
4.  Wang CC, Friedman L, Kennedy GC, Wang H, Kebebew E, Steward DL, Zeiger MA, Westra WH, Wang Y, Khanafshar E, Fellegara G, Rosai J, Livolsi V, Lanman RB. A large multicenter correlation study of thyroid nodule cytopathology and histopathology. Thyroid. 2011;21:243-251.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 265]  [Cited by in RCA: 263]  [Article Influence: 18.8]  [Reference Citation Analysis (0)]
5.  Tollin SR, Mery GM, Jelveh N, Fallon EF, Mikhail M, Blumenfeld W, Perlmutter S. The use of fine-needle aspiration biopsy under ultrasound guidance to assess the risk of malignancy in patients with a multinodular goiter. Thyroid. 2000;10:235-241.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 83]  [Cited by in RCA: 61]  [Article Influence: 2.4]  [Reference Citation Analysis (0)]
6.  Haugen BR, Alexander EK, Bible KC, Doherty GM, Mandel SJ, Nikiforov YE, Pacini F, Randolph GW, Sawka AM, Schlumberger M, Schuff KG, Sherman SI, Sosa JA, Steward DL, Tuttle RM, Wartofsky L. 2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer: The American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer. Thyroid. 2016;26:1-133.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 10769]  [Cited by in RCA: 9746]  [Article Influence: 1082.9]  [Reference Citation Analysis (1)]
7.  Tessler FN, Middleton WD, Grant EG, Hoang JK, Berland LL, Teefey SA, Cronan JJ, Beland MD, Desser TS, Frates MC, Hammers LW, Hamper UM, Langer JE, Reading CC, Scoutt LM, Stavros AT. ACR Thyroid Imaging, Reporting and Data System (TI-RADS): White Paper of the ACR TI-RADS Committee. J Am Coll Radiol. 2017;14:587-595.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 900]  [Cited by in RCA: 1420]  [Article Influence: 177.5]  [Reference Citation Analysis (0)]
8.  Cibas ES, Ali SZ. The 2017 Bethesda System for Reporting Thyroid Cytopathology. Thyroid. 2017;27:1341-1346.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 955]  [Cited by in RCA: 1179]  [Article Influence: 147.4]  [Reference Citation Analysis (0)]
9.  Nikiforova MN, Mercurio S, Wald AI, Barbi de Moura M, Callenberg K, Santana-Santos L, Gooding WE, Yip L, Ferris RL, Nikiforov YE. Analytical performance of the ThyroSeq v3 genomic classifier for cancer diagnosis in thyroid nodules. Cancer. 2018;124:1682-1690.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 198]  [Cited by in RCA: 282]  [Article Influence: 40.3]  [Reference Citation Analysis (0)]
10.  Patel KN, Angell TE, Babiarz J, Barth NM, Blevins T, Duh QY, Ghossein RA, Harrell RM, Huang J, Kennedy GC, Kim SY, Kloos RT, LiVolsi VA, Randolph GW, Sadow PM, Shanik MH, Sosa JA, Traweek ST, Walsh PS, Whitney D, Yeh MW, Ladenson PW. Performance of a Genomic Sequencing Classifier for the Preoperative Diagnosis of Cytologically Indeterminate Thyroid Nodules. JAMA Surg. 2018;153:817-824.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 171]  [Cited by in RCA: 247]  [Article Influence: 41.2]  [Reference Citation Analysis (0)]
11.  Santos MTD, Buzolin AL, Gama RR, Silva ECAD, Dufloth RM, Figueiredo DLA, Carvalho AL. Molecular Classification of Thyroid Nodules with Indeterminate Cytology: Development and Validation of a Highly Sensitive and Specific New miRNA-Based Classifier Test Using Fine-Needle Aspiration Smear Slides. Thyroid. 2018;28:1618-1626.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 26]  [Cited by in RCA: 32]  [Article Influence: 4.6]  [Reference Citation Analysis (0)]
12.  Zarkesh M, Zadeh-Vakili A, Azizi F, Foroughi F, Akhavan MM, Hedayati M. Altered Epigenetic Mechanisms in Thyroid Cancer Subtypes. Mol Diagn Ther. 2018;22:41-56.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 23]  [Cited by in RCA: 28]  [Article Influence: 4.0]  [Reference Citation Analysis (0)]
13.  Ahmed AA, Essa MEA. Potential of epigenetic events in human thyroid cancer. Cancer Genet. 2019;239:13-21.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 10]  [Cited by in RCA: 18]  [Article Influence: 3.0]  [Reference Citation Analysis (0)]
14.  Castillo T JM, Starmans MPA, Arif M, Niessen WJ, Klein S, Bangma CH, Schoots IG, Veenland JF. A Multi-Center, Multi-Vendor Study to Evaluate the Generalizability of a Radiomics Model for Classifying Prostate cancer: High Grade vs. Low Grade. Diagnostics (Basel). 2021;11:369.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 11]  [Cited by in RCA: 33]  [Article Influence: 8.3]  [Reference Citation Analysis (0)]
15.  Ratnani I, Fatima S, Abid MM, Surani Z, Surani S. Evidence-Based Medicine: History, Review, Criticisms, and Pitfalls. Cureus. 2023;15:e35266.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 12]  [Reference Citation Analysis (0)]
16.  Schuijf JD, Lima JAC, Boedeker KL, Takagi H, Tanaka R, Yoshioka K, Arbab-Zadeh A. CT imaging with ultra-high-resolution: Opportunities for cardiovascular imaging in clinical practice. J Cardiovasc Comput Tomogr. 2022;16:388-396.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 27]  [Reference Citation Analysis (0)]
17.  Buchbinder BR. Functional magnetic resonance imaging. Handb Clin Neurol. 2016;135:61-92.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 34]  [Cited by in RCA: 43]  [Article Influence: 4.8]  [Reference Citation Analysis (0)]
18.  Finn ES, Poldrack RA, Shine JM. Functional neuroimaging as a catalyst for integrated neuroscience. Nature. 2023;623:263-273.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 61]  [Cited by in RCA: 37]  [Article Influence: 18.5]  [Reference Citation Analysis (0)]
19.  Zhao L. Advances in functional magnetic resonance imaging-based brain function mapping: a deep learning perspective. Psychoradiology. 2025;5:kkaf007.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
20.  Nagaraja N. Diffusion weighted imaging in acute ischemic stroke: A review of its interpretation pitfalls and advanced diffusion imaging application. J Neurol Sci. 2021;425:117435.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 8]  [Cited by in RCA: 23]  [Article Influence: 5.8]  [Reference Citation Analysis (0)]
21.  Wang S, Pan J, Zhang X, Li Y, Liu W, Lin R, Wang X, Kang D, Li Z, Huang F, Chen L, Chen J. Towards next-generation diagnostic pathology: AI-empowered label-free multiphoton microscopy. Light Sci Appl. 2024;13:254.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 7]  [Reference Citation Analysis (0)]
22.  Khalid N, Zubair M, Mehmood MQ, Massoud Y. Emerging paradigms in microwave imaging technology for biomedical applications: unleashing the power of artificial intelligence. Npj Imaging. 2024;2:13.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 3]  [Reference Citation Analysis (0)]
23.  Łoginoff J, Majos A, Elgalal M. The Evolution of Custom Subperiosteal Implants for Treatment of Partial or Complete Edentulism in Patients with Severe Alveolar Ridge Atrophy. J Clin Med. 2024;13:3582.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 8]  [Reference Citation Analysis (0)]
24.  Margulis AR. Recent advances in clinical diagnostic radiology. 4. Interventional diagnostic roentgenology. Adv Surg. 1971;5:88-102.  [PubMed]  [DOI]
25.  Harvey CJ, Pilcher JM, Eckersley RJ, Blomley MJ, Cosgrove DO. Advances in ultrasound. Clin Radiol. 2002;57:157-177.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 141]  [Cited by in RCA: 115]  [Article Influence: 5.0]  [Reference Citation Analysis (0)]
26.  Golemati S, Cokkinos DD. Recent advances in vascular ultrasound imaging technology and their clinical implications. Ultrasonics. 2022;119:106599.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 9]  [Cited by in RCA: 42]  [Article Influence: 14.0]  [Reference Citation Analysis (0)]
27.  Massey A, Stewart J, Smith C, Parvini C, McCormick M, Do K, Cartagena-Rivera AX. Mechanical properties of human tumour tissues and their implications for cancer development. Nat Rev Phys. 2024;6:269-282.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 30]  [Reference Citation Analysis (0)]
28.  Blake OO, Ramsook R, Faulkner DR, Iyare UC. The Effect of Effective Pressure on the Relationship Between Static and Dynamic Young’s Moduli and Poisson’s Ratio of Naparima Hill Formation Mudstones. Rock Mech Rock Eng. 2020;53:3761-3778.  [PubMed]  [DOI]  [Full Text]
29.  Wang Y, Wang Q, Pei X, Mei S, Liu J.   Optimizing Pointwise Convolutions on Multi-core DSPs. In: Tari Z, Li K, Wu H. Algorithms and Architectures for Parallel Processing. ICA3PP 2023. Lecture Notes in Computer Science. Singapore: Springer, 2024.  [PubMed]  [DOI]  [Full Text]
30.  Liu Z, Ni S, Yang C, Sun W, Huang D, Su H, Shu J, Qin N. Axillary lymph node metastasis prediction by contrast-enhanced computed tomography images for breast cancer patients based on deep learning. Comput Biol Med. 2021;136:104715.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 4]  [Cited by in RCA: 18]  [Article Influence: 4.5]  [Reference Citation Analysis (0)]
31.  Zhang X, Liang M, Yang Z, Zheng C, Wu J, Ou B, Li H, Wu X, Luo B, Shen J. Deep Learning-Based Radiomics of B-Mode Ultrasonography and Shear-Wave Elastography: Improved Performance in Breast Mass Classification. Front Oncol. 2020;10:1621.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 44]  [Cited by in RCA: 40]  [Article Influence: 8.0]  [Reference Citation Analysis (0)]
32.  Çetin Tunçez H, Murat Koç A, Hilal Adıbelli Z, Zeynep Arslan F, Argon A, Yücel Oğuzdoğan G. Diagnostic Efficacy of Ultrasonography, Doppler Ultrasonography and Elastography in the Evaluation of Suspected Malignant Lymph Nodes. J Ultrason. 2023;23:1-9.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 2]  [Reference Citation Analysis (0)]
33.  Pu C, Wu T, Wang Q, Yang Y, Zhang K. Feasibility of novel intraoperative navigation for anatomical liver resection using real-time virtual sonography combined with indocyanine green fluorescent imaging technology. Biosci Trends. 2024;17:484-490.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 2]  [Reference Citation Analysis (0)]
34.  Ippolito D, Maino C, Pecorelli A, Salemi I, Gandola D, Riva L, Talei Franzesi C, Sironi S. Application of low-dose CT combined with model-based iterative reconstruction algorithm in oncologic patients during follow-up: dose reduction and image quality. Br J Radiol. 2021;94:20201223.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 6]  [Cited by in RCA: 6]  [Article Influence: 1.5]  [Reference Citation Analysis (0)]
35.  Chi J, Xu D, Yin S, Li M, Shen L, Ding N, Chen X, Ji Y. Reducing the radiation dose of pediatric paranasal sinus CT using an ultralow tube voltage (70 kVp) combined with iterative reconstruction: Feasibility and image quality. Medicine (Baltimore). 2020;99:e21886.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 5]  [Cited by in RCA: 7]  [Article Influence: 1.4]  [Reference Citation Analysis (0)]
36.  Bosch de Basea M, Thierry-Chef I, Harbron R, Hauptmann M, Byrnes G, Bernier MO, Le Cornet L, Dabin J, Ferro G, Istad TS, Jahnen A, Lee C, Maccia C, Malchair F, Olerud H, Simon SL, Figuerola J, Peiro A, Engels H, Johansen C, Blettner M, Kaijser M, Kjaerheim K, Berrington de Gonzalez A, Journy N, Meulepas JM, Moissonnier M, Nordenskjold A, Pokora R, Ronckers C, Schüz J, Kesminiene A, Cardis E. Risk of hematological malignancies from CT radiation exposure in children, adolescents and young adults. Nat Med. 2023;29:3111-3119.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 126]  [Cited by in RCA: 128]  [Article Influence: 64.0]  [Reference Citation Analysis (0)]
37.  Matsunaga Y, Haba T, Kobayashi M, Suzuki S, Asada Y, Chida K. Assessment of fetal radiation exposure in pregnant women undergoing computed tomography and rotational angiography examinations for pelvic trauma. Radiat Prot Dosimetry. 2024;200:580-587.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
38.  Zhu S, Zhang B, Tian Q, Li A, Liu Z, Hou W, Zhao W, Huang X, Xiao Y, Wang Y, Wang R, Li Y, Yang J, Jin C. Reduced-dose deep learning iterative reconstruction for abdominal computed tomography with low tube voltage and tube current. BMC Med Inform Decis Mak. 2024;24:389.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
39.  Solomon DZ, Ayalew B, Dellie ST, Admasie D. Justification and Optimization Principles of ALARA in Pediatric CT at a Teaching Hospital in Ethiopia. Ethiop J Health Sci. 2020;30:761-766.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1]  [Cited by in RCA: 5]  [Article Influence: 1.3]  [Reference Citation Analysis (0)]
40.  Berrington de Gonzalez A, Pasqual E, Veiga L. Epidemiological studies of CT scans and cancer risk: the state of the science. Br J Radiol. 2021;94:20210471.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 5]  [Cited by in RCA: 32]  [Article Influence: 8.0]  [Reference Citation Analysis (0)]
41.  Brenner DJ, Doll R, Goodhead DT, Hall EJ, Land CE, Little JB, Lubin JH, Preston DL, Preston RJ, Puskin JS, Ron E, Sachs RK, Samet JM, Setlow RB, Zaider M. Cancer risks attributable to low doses of ionizing radiation: assessing what we really know. Proc Natl Acad Sci U S A. 2003;100:13761-13766.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1180]  [Cited by in RCA: 1130]  [Article Influence: 51.4]  [Reference Citation Analysis (0)]
42.  Mullenders L, Atkinson M, Paretzke H, Sabatier L, Bouffler S. Assessing cancer risks of low-dose radiation. Nat Rev Cancer. 2009;9:596-604.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 163]  [Cited by in RCA: 160]  [Article Influence: 10.0]  [Reference Citation Analysis (0)]
43.  Hauptmann M, Daniels RD, Cardis E, Cullings HM, Kendall G, Laurier D, Linet MS, Little MP, Lubin JH, Preston DL, Richardson DB, Stram DO, Thierry-Chef I, Schubauer-Berigan MK, Gilbert ES, Berrington de Gonzalez A. Epidemiological Studies of Low-Dose Ionizing Radiation and Cancer: Summary Bias Assessment and Meta-Analysis. J Natl Cancer Inst Monogr. 2020;2020:188-200.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 92]  [Cited by in RCA: 116]  [Article Influence: 23.2]  [Reference Citation Analysis (0)]
44.  Kelaranta A, Mäkelä T, Kaasalainen T, Kortesniemi M. Fetal radiation dose in three common CT examinations during pregnancy - Monte Carlo study. Phys Med. 2017;43:199-206.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 10]  [Cited by in RCA: 26]  [Article Influence: 3.3]  [Reference Citation Analysis (0)]
45.  Patro SN, Chakraborty S, Sheikh A. The use of adaptive statistical iterative reconstruction (ASiR) technique in evaluation of patients with cervical spine trauma: impact on radiation dose reduction and image quality. Br J Radiol. 2016;89:20150082.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 13]  [Cited by in RCA: 17]  [Article Influence: 1.9]  [Reference Citation Analysis (0)]
46.  Wang J, Sui X, Zhao R, Du H, Wang J, Wang Y, Qin R, Lu X, Ma Z, Xu Y, Jin Z, Song L, Song W. Value of deep learning reconstruction of chest low-dose CT for image quality improvement and lung parenchyma assessment on lung window. Eur Radiol. 2024;34:1053-1064.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2]  [Cited by in RCA: 7]  [Article Influence: 7.0]  [Reference Citation Analysis (0)]
47.  Jiang JM, Miao L, Liang X, Liu ZH, Zhang L, Li M. The Value of Deep Learning Image Reconstruction in Improving the Quality of Low-Dose Chest CT Images. Diagnostics (Basel). 2022;12:2560.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 7]  [Reference Citation Analysis (0)]
48.  Jo GD, Ahn C, Hong JH, Kim DS, Park J, Kim H, Kim JH, Goo JM, Nam JG. 75% radiation dose reduction using deep learning reconstruction on low-dose chest CT. BMC Med Imaging. 2023;23:121.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 7]  [Reference Citation Analysis (0)]
49.  Modi L, Sun W, Shafizadeh N, Negron R, Yee-Chang M, Zhou F, Simsir A, Sheth S, Brandler TC. Does a higher American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) score forecast an increased risk of malignancy? A correlation study of ACR TI-RADS with FNA cytology in the evaluation of thyroid nodules. Cancer Cytopathol. 2020;128:470-481.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 11]  [Cited by in RCA: 27]  [Article Influence: 5.4]  [Reference Citation Analysis (0)]
50.  Cui B, Wu Z, Yu B, Li Y, Liu H. The Diagnostic Value of BRAF V600E Gene Detection Combined with DNA Ploidy Analysis in Thyroid Cancer. Discov Med. 2023;35:1064-1070.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1]  [Cited by in RCA: 2]  [Article Influence: 1.0]  [Reference Citation Analysis (0)]
51.  Cancer Genome Atlas Research Network. Integrated genomic characterization of papillary thyroid carcinoma. Cell. 2014;159:676-690.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1650]  [Cited by in RCA: 2197]  [Article Influence: 219.7]  [Reference Citation Analysis (0)]
52.  Vardarli I, Tan S, Görges R, Krämer BK, Herrmann K, Brochhausen C. Diagnostic accuracy of Afirma gene expression classifier, Afirma gene sequencing classifier, ThyroSeq v2 and ThyroSeq v3 for indeterminate (Bethesda III and IV) thyroid nodules: a meta-analysis. Endocr Connect. 2024;13:e240170.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 4]  [Reference Citation Analysis (0)]
53.  Silaghi CA, Lozovanu V, Georgescu CE, Georgescu RD, Susman S, Năsui BA, Dobrean A, Silaghi H. Thyroseq v3, Afirma GSC, and microRNA Panels Versus Previous Molecular Tests in the Preoperative Diagnosis of Indeterminate Thyroid Nodules: A Systematic Review and Meta-Analysis. Front Endocrinol (Lausanne). 2021;12:649522.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 22]  [Cited by in RCA: 47]  [Article Influence: 11.8]  [Reference Citation Analysis (0)]
54.  Wu ST, Voltoline R, Benites RL, de Campos BM, de Souza JPSS, Ghizoni E. Interactive mining of neural pathways to preoperative neurosurgical planning. Comput Biol Med. 2025;184:109334.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
55.  Wu W, Poser BA, Douaud G, Frost R, In MH, Speck O, Koopmans PJ, Miller KL. High-resolution diffusion MRI at 7T using a three-dimensional multi-slab acquisition. Neuroimage. 2016;143:1-14.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 42]  [Cited by in RCA: 52]  [Article Influence: 5.8]  [Reference Citation Analysis (0)]
56.  Niaz MR, Ridwan AR, Wu Y; Alzheimer's Disease Neuroimaging Initiative, Bennett DA, Arfanakis K. Development and evaluation of a high resolution 0.5mm isotropic T1-weighted template of the older adult brain. Neuroimage. 2022;248:118869.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 8]  [Cited by in RCA: 9]  [Article Influence: 3.0]  [Reference Citation Analysis (0)]
57.  Menon V, Brash DE. Next-generation sequencing methodologies to detect low-frequency mutations: "Catch me if you can". Mutat Res Rev Mutat Res. 2023;792:108471.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 6]  [Cited by in RCA: 17]  [Article Influence: 8.5]  [Reference Citation Analysis (0)]
58.  Murugan AK, Kannan S, Alzahrani AS. TERT promoter mutations in gliomas: Molecular roles in tumorigenesis, metastasis, diagnosis, prognosis, therapeutic targeting, and drug resistance. Biochim Biophys Acta Rev Cancer. 2025;1880:189243.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2]  [Cited by in RCA: 3]  [Article Influence: 3.0]  [Reference Citation Analysis (0)]
59.  Marczyk VR, Maia AL, Goemann IM. Distinct transcriptional and prognostic impacts of TERT promoter mutations C228T and C250T in papillary thyroid carcinoma. Endocr Relat Cancer. 2024;31:e240058.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 3]  [Reference Citation Analysis (0)]
60.  Sun Z, Du M, Wu X, Tao R, Sun P, Zheng S, Zhang Z, Zhou D, Zhao X, Yang P. Rapid diagnosis of TERT promoter mutation using Terahertz absorption spectroscopy in glioblastoma. Sci Rep. 2025;15:18480.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
61.  Ganeshan B, Miles KA. Quantifying tumour heterogeneity with CT. Cancer Imaging. 2013;13:140-149.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 243]  [Cited by in RCA: 296]  [Article Influence: 24.7]  [Reference Citation Analysis (0)]
62.  Ruan Z, Mei N, Lu Y, Xiong J, Li X, Zheng W, Liu L, Yin B. A Comparative and Summative Study of Radiomics-based Overall Survival Prediction in Glioblastoma Patients. J Comput Assist Tomogr. 2022;46:470-479.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 7]  [Reference Citation Analysis (0)]
63.  Ambrozkiewicz F, Trailin A, Červenková L, Vaclavikova R, Hanicinec V, Allah MAO, Palek R, Třeška V, Daum O, Tonar Z, Liška V, Hemminki K. CTNNB1 mutations, TERT polymorphism and CD8+ cell densities in resected hepatocellular carcinoma are associated with longer time to recurrence. BMC Cancer. 2022;22:884.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1]  [Cited by in RCA: 18]  [Article Influence: 6.0]  [Reference Citation Analysis (0)]
64.  Maloberti T, Repaci A, Poppi L, Di Paola FJ, Calafato G, Coluccelli S, Carosi F, Colapinto A, Colombero S, Credi G, Tallini G, Pantaleo MA, Nannini M, de Biase D. Exploring the role of TERT in thyroid Cancer: A systematic review. Crit Rev Oncol Hematol. 2025;213:104792.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1]  [Cited by in RCA: 1]  [Article Influence: 1.0]  [Reference Citation Analysis (0)]
65.  Haag JG, Wolsky RJ, Moroney MR, Sheren J, Sheeder J, Bitler BG, Corr BR. Clinicopathologic Evaluation of CTNNB1 Mutations in High-Intermediate Risk Endometrial Endometrioid Carcinoma. Int J Gynecol Pathol. 2023;42:43-53.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 3]  [Cited by in RCA: 4]  [Article Influence: 2.0]  [Reference Citation Analysis (0)]
66.  Zhao B. Understanding Sources of Variation to Improve the Reproducibility of Radiomics. Front Oncol. 2021;11:633176.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 42]  [Cited by in RCA: 80]  [Article Influence: 20.0]  [Reference Citation Analysis (0)]
67.  Feng C, Zhang L, Chang X, Qin D, Zhang T. Regulation of post-translational modification of PD-L1 and advances in tumor immunotherapy. Front Immunol. 2023;14:1230135.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 15]  [Cited by in RCA: 24]  [Article Influence: 12.0]  [Reference Citation Analysis (0)]
68.  Gao Y, Fan S, Sun X, Li J, Dai Y, Li H, Ma H, Xu Y, Lv L. Oncometabolite fumarate facilitates PD-L1 expression and immune evasion in clear cell renal cell carcinoma. Cell Death Dis. 2025;16:432.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 3]  [Cited by in RCA: 2]  [Article Influence: 2.0]  [Reference Citation Analysis (0)]
69.  Li B, Bao F, Hou Y, Li F, Li H, Deng Y, Dai Q. Tissue characterization at an enhanced resolution across spatial omics platforms with deep generative model. Nat Commun. 2024;15:6541.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 8]  [Reference Citation Analysis (0)]
70.  Wang H, Zhao S, Yao J, Yu X, Xu D. Factors influencing extrathyroidal extension of papillary thyroid cancer and evaluation of ultrasonography for its diagnosis: a retrospective analysis. Sci Rep. 2023;13:18344.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 5]  [Reference Citation Analysis (0)]
71.  Zhou J, Song Y, Zhan W, Wei X, Zhang S, Zhang R, Gu Y, Chen X, Shi L, Luo X, Yang L, Li Q, Bai B, Ye X, Zhai H, Zhang H, Jia X, Dong Y, Zhang J, Yang Z, Zhang H, Zheng Y, Xu W, Lai L, Yin L; Superficial Organ and Vascular Ultrasound Group of the Society of Ultrasound in Medicine of Chinese Medical Association;  Chinese Artificial Intelligence Alliance for Thyroid and Breast Ultrasound. Thyroid imaging reporting and data system (TIRADS) for ultrasound features of nodules: multicentric retrospective study in China. Endocrine. 2021;72:157-170.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 15]  [Cited by in RCA: 28]  [Article Influence: 7.0]  [Reference Citation Analysis (0)]
72.  Alexander EK, Marqusee E, Orcutt J, Benson CB, Frates MC, Doubilet PM, Cibas ES, Atri A. Thyroid nodule shape and prediction of malignancy. Thyroid. 2004;14:953-958.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 100]  [Cited by in RCA: 95]  [Article Influence: 4.5]  [Reference Citation Analysis (0)]
73.  Song RY, Kim HS, Kang KH. Minimal extrathyroidal extension is associated with lymph node metastasis in single papillary thyroid microcarcinoma: a retrospective analysis of 814 patients. World J Surg Oncol. 2022;20:170.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 16]  [Reference Citation Analysis (0)]
74.  Cao M, Yu T, Miao X, Wu Z, Wang W. The preferred surgical choice for intermediate-risk papillary thyroid cancer: total thyroidectomy or lobectomy? A systematic review and meta-analysis. Int J Surg. 2024;110:5087-5100.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 5]  [Cited by in RCA: 8]  [Article Influence: 8.0]  [Reference Citation Analysis (0)]
75.  Gong H, Jiang Y, Su A. Total thyroidectomy versus less-than-total thyroidectomy for papillary thyroid carcinoma of isthmus: a systematic review and meta-analysis. Gland Surg. 2023;12:1525-1540.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 5]  [Reference Citation Analysis (0)]
76.  Lee EJ, Song KH, Kim DL, Jang YM, Hwang TS, Kim SK. The BRAF(V600E) mutation is associated with malignant ultrasonographic features in thyroid nodules. Clin Endocrinol (Oxf). 2011;75:844-850.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 23]  [Cited by in RCA: 28]  [Article Influence: 2.0]  [Reference Citation Analysis (0)]
77.  Hong X, Li J, Duan S, You Y. Retrospective study of BRAF(V600E) mutation and CT features of papillary thyroid carcinoma. PeerJ. 2024;12:e16810.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
78.  Howell GM, Hodak SP, Yip L. RAS mutations in thyroid cancer. Oncologist. 2013;18:926-932.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 98]  [Cited by in RCA: 141]  [Article Influence: 11.8]  [Reference Citation Analysis (0)]
79.  Burge RA, Hobbs GA. Not all RAS mutations are equal: A detailed review of the functional diversity of RAS hot spot mutations. Adv Cancer Res. 2022;153:29-61.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 18]  [Reference Citation Analysis (0)]
80.  Liu SZ, Yang SH, Ye M, Fu BJ, Lv FJ, Chu ZG. Bubble-like lucency in pulmonary ground glass nodules on computed tomography: a specific pattern of air-containing space for diagnosing neoplastic lesions. Cancer Imaging. 2024;24:47.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 4]  [Reference Citation Analysis (0)]
81.  Sutton RT, Pincock D, Baumgart DC, Sadowski DC, Fedorak RN, Kroeker KI. An overview of clinical decision support systems: benefits, risks, and strategies for success. NPJ Digit Med. 2020;3:17.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1236]  [Cited by in RCA: 1170]  [Article Influence: 234.0]  [Reference Citation Analysis (0)]
82.  Jie Z, Zhiying Z, Li L. A meta-analysis of Watson for Oncology in clinical application. Sci Rep. 2021;11:5792.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 13]  [Cited by in RCA: 64]  [Article Influence: 16.0]  [Reference Citation Analysis (0)]
83.  Byrd TF 4th, Southwell B, Ravishankar A, Tran T, Kc A, Phelan T, Melton-Meaux GB, Usher MG, Scheppmann D, Switzer S, Simon G, Tignanelli CJ. Validation of a Proprietary Deterioration Index Model and Performance in Hospitalized Adults. JAMA Netw Open. 2023;6:e2324176.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 14]  [Reference Citation Analysis (0)]
84.  Jones C, Thornton J, Wyatt JC. Artificial intelligence and clinical decision support: clinicians' perspectives on trust, trustworthiness, and liability. Med Law Rev. 2023;31:501-520.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 6]  [Cited by in RCA: 34]  [Article Influence: 17.0]  [Reference Citation Analysis (0)]
85.  Nakauchi M, Court CM, Tang LH, Gönen M, Janjigian YY, Maron SB, Molena D, Coit DG, Brennan MF, Strong VE. Validation of the Memorial Sloan Kettering Gastric Cancer Post-Resection Survival Nomogram: Does It Stand the Test of Time? J Am Coll Surg. 2022;235:294-304.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 6]  [Cited by in RCA: 7]  [Article Influence: 2.3]  [Reference Citation Analysis (0)]
86.  Ye J, Hai J, Song J, Wang Z. Multimodal Data Hybrid Fusion and Natural Language Processing for Clinical Prediction Models. AMIA Jt Summits Transl Sci Proc. 2024;2024:191-200.  [PubMed]  [DOI]
87.  Paolucci I, Lin YM, Albuquerque Marques Silva J, Brock KK, Odisio BC. Bayesian parametric models for survival prediction in medical applications. BMC Med Res Methodol. 2023;23:250.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 5]  [Cited by in RCA: 3]  [Article Influence: 1.5]  [Reference Citation Analysis (0)]
88.  Zwanenburg A, Vallières M, Abdalah MA, Aerts HJWL, Andrearczyk V, Apte A, Ashrafinia S, Bakas S, Beukinga RJ, Boellaard R, Bogowicz M, Boldrini L, Buvat I, Cook GJR, Davatzikos C, Depeursinge A, Desseroit MC, Dinapoli N, Dinh CV, Echegaray S, El Naqa I, Fedorov AY, Gatta R, Gillies RJ, Goh V, Götz M, Guckenberger M, Ha SM, Hatt M, Isensee F, Lambin P, Leger S, Leijenaar RTH, Lenkowicz J, Lippert F, Losnegård A, Maier-Hein KH, Morin O, Müller H, Napel S, Nioche C, Orlhac F, Pati S, Pfaehler EAG, Rahmim A, Rao AUK, Scherer J, Siddique MM, Sijtsema NM, Socarras Fernandez J, Spezi E, Steenbakkers RJHM, Tanadini-Lang S, Thorwarth D, Troost EGC, Upadhaya T, Valentini V, van Dijk LV, van Griethuysen J, van Velden FHP, Whybra P, Richter C, Löck S. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology. 2020;295:328-338.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2090]  [Cited by in RCA: 2354]  [Article Influence: 470.8]  [Reference Citation Analysis (0)]
89.  Printz C. Immunotherapy combination shows longer progression-free survival than chemotherapy alone for patients with a rare type of metastatic colorectal cancer. CA Cancer J Clin. 2025;75:87-89.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
90.  Xu H, Zhang Y, Wu H, Zhou N, Li X, Pineda JP, Zhu Y, Fu H, Ying M, Yang S, Bao J, Yang L, Zhang B, Guo L, Sun L, Lu F, Wang H, Huang Y, Zhu T, Wang X, Wei Q, Sheng C, Qu S, Lv Z, Xu D, Li Q, Dong Y, Qin J, Cheng T, Xing M. High Diagnostic Accuracy of Epigenetic Imprinting Biomarkers in Thyroid Nodules. J Clin Oncol. 2023;41:1296-1306.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 18]  [Cited by in RCA: 14]  [Article Influence: 7.0]  [Reference Citation Analysis (0)]
91.  Zhou L, Wang W, Wei H, Song P, Li Z, Cheng L, Lei P, Chen Q, Liu Z, Ye H, Cai Q. Clinical application of 3D Slicer combined with Sina/MosoCam multimodal system in preoperative planning of brain lesions surgery. Sci Rep. 2022;12:19258.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 8]  [Reference Citation Analysis (0)]
92.  Cè M, Irmici G, Foschini C, Danesini GM, Falsitta LV, Serio ML, Fontana A, Martinenghi C, Oliva G, Cellina M. Artificial Intelligence in Brain Tumor Imaging: A Step toward Personalized Medicine. Curr Oncol. 2023;30:2673-2701.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 37]  [Reference Citation Analysis (0)]
93.  Zhang P, Luo H, Zhu W, Yang J, Zeng N, Fan Y, Wen S, Xiang N, Jia F, Fang C. Real-time navigation for laparoscopic hepatectomy using image fusion of preoperative 3D surgical plan and intraoperative indocyanine green fluorescence imaging. Surg Endosc. 2020;34:3449-3459.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 74]  [Cited by in RCA: 66]  [Article Influence: 13.2]  [Reference Citation Analysis (0)]
94.  Bond C, Santiago-Ruiz AN, Tang Q, Lakadamyali M. Technological advances in super-resolution microscopy to study cellular processes. Mol Cell. 2022;82:315-332.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 58]  [Cited by in RCA: 59]  [Article Influence: 19.7]  [Reference Citation Analysis (0)]
95.  Singh D, Tewari M, Singh S, Narayan G. Revisiting the role of TRAIL/TRAIL-R in cancer biology and therapy. Future Oncol. 2021;17:581-596.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 7]  [Cited by in RCA: 31]  [Article Influence: 7.8]  [Reference Citation Analysis (0)]
96.  Wu Z, Hua P, Chen X, Lei J, Zhang L, Zhang P. A study on the prediction of targeted therapy efficacy in advanced lung adenocarcinoma patients with EGFR mutations using CT-based delta-radiomics model. Front Med (Lausanne). 2025;12:1599206.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 1]  [Reference Citation Analysis (0)]
97.  Wang Y, Guan Q, Lao I, Wang L, Wu Y, Li D, Ji Q, Wang Y, Zhu Y, Lu H, Xiang J. Using deep convolutional neural networks for multi-classification of thyroid tumor by histopathology: a large-scale pilot study. Ann Transl Med. 2019;7:468.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 26]  [Cited by in RCA: 50]  [Article Influence: 8.3]  [Reference Citation Analysis (0)]
98.  Domingo RP, Ogden LL, Been LC, Kennedy GC, Traweek ST. Identification of parathyroid tissue in thyroid fine-needle aspiration: A combined approach using cytology, immunohistochemical, and molecular methods. Diagn Cytopathol. 2017;45:526-532.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 13]  [Cited by in RCA: 15]  [Article Influence: 1.9]  [Reference Citation Analysis (0)]
99.  Kim KH, Migliozzi S, Koo H, Hong JH, Park SM, Kim S, Kwon HJ, Ha S, Garofano L, Oh YT, D'Angelo F, Kim CI, Kim S, Lee JY, Kim J, Hong J, Jang EH, Mathon B, Di Stefano AL, Bielle F, Laurenge A, Nesvizhskii AI, Hur EM, Yin J, Shi B, Kim Y, Moon KS, Kwon JT, Lee SH, Lee SH, Gwak HS, Lasorella A, Yoo H, Sanson M, Sa JK, Park CK, Nam DH, Iavarone A, Park JB. Integrated proteogenomic characterization of glioblastoma evolution. Cancer Cell. 2024;42:358-377.e8.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1]  [Cited by in RCA: 36]  [Article Influence: 36.0]  [Reference Citation Analysis (0)]
100.  Schulte J, Caliebe A, Marciano M, Neuschwander P, Seiberle I, Scheurer E, Schulz I. DEPArray™ single-cell technology: A validation study for forensic applications. Forensic Sci Int Genet. 2024;70:103026.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 6]  [Reference Citation Analysis (0)]
101.  Hui Z, Zhang J, Ren Y, Li X, Yan C, Yu W, Wang T, Xiao S, Chen Y, Zhang R, Wei F, You J, Ren X. Single-cell profiling of immune cells after neoadjuvant pembrolizumab and chemotherapy in IIIA non-small cell lung cancer (NSCLC). Cell Death Dis. 2022;13:607.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 2]  [Cited by in RCA: 45]  [Article Influence: 15.0]  [Reference Citation Analysis (0)]
102.  Liu Y, Yang M, Deng Y, Su G, Enninful A, Guo CC, Tebaldi T, Zhang D, Kim D, Bai Z, Norris E, Pan A, Li J, Xiao Y, Halene S, Fan R. High-Spatial-Resolution Multi-Omics Sequencing via Deterministic Barcoding in Tissue. Cell. 2020;183:1665-1681.e18.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 182]  [Cited by in RCA: 539]  [Article Influence: 107.8]  [Reference Citation Analysis (0)]
103.  Francies HE, Barthorpe A, McLaren-Douglas A, Barendt WJ, Garnett MJ. Drug Sensitivity Assays of Human Cancer Organoid Cultures. Methods Mol Biol. 2019;1576:339-351.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 49]  [Cited by in RCA: 56]  [Article Influence: 9.3]  [Reference Citation Analysis (0)]