Domain gaps are significant obstacles to the clinical implementation of machine learning (ML) solutions for medical image analysis. Although current research emphasizes new training methods and network architectures, the specific impact of prevalence shifts on algorithms in real-world applications is often overlooked. Differences in class frequencies between development and deployment data are crucial, particularly for the widespread adoption of artificial intelligence (AI), as disease prevalence can vary greatly across different times and locations. Our contribution is threefold. Based on a diverse set of 30 medical classification tasks (1) we demonstrate that lack of prevalence shift handling can have severe consequences on the quality of calibration, decision threshold, and performance assessment. Furthermore, (2) we show that prevalences can be accurately and reliably estimated in a data-driven manner. Finally, (3) we propose a new workflow for prevalence-aware image classification that uses estimated deployment prevalences to adjust a trained classifier to a new environment, without requiring additional annotated deployment data. Comprehensive experiments indicate that our proposed approach could contribute to generating better classifier decisions and more reliable performance estimates compared to current practice.

For patients suffering from differentiated thyroid cancer (DTC), several clinical, laboratory, and pathological features (including patient age, tumor size, extrathyroidal extension, or serum thyroglobulin levels) are currently used to identify recurrence risk. Validation and potential adjustment of their individual and combined prognostic values using a large patient cohort with several years of follow-up might improve the correct identification of patients at risk.

In this retrospective study, we developed an XGBoost model using clinical and biomarker features for accurate DTC recurrence prediction using a cohort of 1228 consecutive patients (965 papillary, and 263 follicular) that were treated at the Department of Nuclear Medicine at University Hospital Augsburg between 1976 and 2010. The dataset was split into 982 patients for model training, and 246 for independent testing. From the 982 patients, 200 different random combinations of 785 training and 197 validation patients were conducted. To identify critical risk factors and understand the model's decision-making process, we conducted Shapely Additive exPlanations (SHAP) analysis.

The XGBoost model achieved an AUROC of 0.84 (95 % CI: 0.84-0.86; SD: 0.08), sensitivity of 0.79 (95 % CI: 0.77-0.81; SD: 0.17), and specificity of 0.78 (95 % CI: 0.77-0.79; SD: 0.04) on the validation datasets, and an AUROC of 0.88 (sensitivity 0.83, specificity 0.80) on the independent test set. Tumor size, maximal thyroglobulin values within six months after thyroidectomy, and maximal thyroglobulin antibody levels within 12 to 24 months after thyroidectomy were the most important factors. SHAP dependence plots suggested new recurrence risk thresholds for a tumor size of 25 mm, maximal serum thyroglobulin levels of 3 and 10 ng/mL, respectively, and maximal thyroglobulin antibody levels of 120 IU/mL.

Our XGBoost model, supported by SHAP analysis empowers clinicians with interpretable insights and defined risk thresholds and could facilitate informed decision-making and patient-centric care.

Recent developments in artificial intelligence (AI) may significantly alter physiological research and healthcare delivery. Whereas AI applications in medicine have historically been trained for specific tasks, recent technological advances have produced models trained on more diverse datasets with much higher parameter counts. These new, "foundation" models raise the possibility that more flexible AI tools can be applied to a wider set of healthcare tasks than in the past. This review describes how these newer models differ from conventional task-specific AI, which relies heavily on focused datasets and narrow, specific applications. By examining the integration of AI into diagnostic tools, personalized treatment strategies, biomedical research, and healthcare administration, we highlight how these newer models are revolutionizing predictive healthcare analytics and operational workflows. In addition, we address ethical and practical considerations associated with the use of foundation models by highlighting emerging trends, calling for changes to existing guidelines, and emphasizing the importance of aligning AI with clinical goals to ensure its responsible and effective use.

Machine learning (ML) models are increasingly being applied to diagnose and predict disease, but face technical challenges such as population drift, where the training and real-world deployed data distributions differ. This phenomenon can degrade model performance, risking incorrect diagnoses. Current detection methods are limited: not directly measuring population drift and often requiring ground truth labels for new patient data. Here, we propose using a one-class support vector machine (OCSVM) to detect population drift. We trained a OCSVM on the Wisconsin Breast Cancer dataset and tested its ability to detect population drift on simulated data. Simulated data was offset at 0.4 standard deviations of the minimum and maximum values of the radius_mean variable, at three noise levels: 5%, 10% and 30% of the standard deviation; 10,000 records per noise level. We hypothesised that increased noise would correlate with more OCSVM-detected inliers, indicating a sensitivity to population drift. As noise increased, more inliers were detected: 5% (27 inliers), 10% (486), and 30% (851). Therefore, this approach could effectively alert to population drift, supporting safe ML diagnostics adoption. Future research should explore OCSVM monitoring on real-world data, enhance model transparency, investigate complementary statistical and ML methods, and extend applications to other data types.

Generative artificial intelligence (AI) has emerged as a transformative force in various fields, including anatomic pathology, where it offers the potential to significantly enhance diagnostic accuracy, workflow efficiency, and research capabilities.

To explore the applications, benefits, and challenges of generative AI in anatomic pathology, with a focus on its impact on diagnostic processes, workflow efficiency, education, and research.

A comprehensive review of current literature and recent advancements in the application of generative AI within anatomic pathology, categorized into unimodal and multimodal applications, and evaluated for clinical utility, ethical considerations, and future potential.

Generative AI demonstrates significant promise in various domains of anatomic pathology, including diagnostic accuracy enhanced through AI-driven image analysis, virtual staining, and synthetic data generation; workflow efficiency, with potential for improvement by automating routine tasks, quality control, and reflex testing; education and research, facilitated by AI-generated educational content, synthetic histology images, and advanced data analysis methods; and clinical integration, with preliminary surveys indicating cautious optimism for nondiagnostic AI tasks and growing engagement in academic settings. Ethical and practical challenges require rigorous validation, prompt engineering, federated learning, and synthetic data generation to help ensure trustworthy, reliable, and unbiased AI applications. Generative AI can potentially revolutionize anatomic pathology, enhancing diagnostic accuracy, improving workflow efficiency, and advancing education and research. Successful integration into clinical practice will require continued interdisciplinary collaboration, careful validation, and adherence to ethical standards to ensure the benefits of AI are realized while maintaining the highest standards of patient care.

The supply of safe pipeline medical devices is of paramount importance. Opportunities exist to transform reusable medical devices for improved processing that meets diverse patient needs. There is increased interest in multi-actor hub frameworks to meet innovation challenges globally. The purpose of this scoping paper was to identify critical decontamination and sterilization needs for the medtech and pharmaceutical sectors with a focus on understanding how to effectively use the Penta helix hub framework that combines academia, industry, healthcare, policy-makers/regulators and patients/society. A PRISMA scoping review of PubMed publications was conducted over the period 2010 to January 2025. Thirty of the 124 'helix hub' papers addressed innovation where only 3 of 16 healthcare-focused helices used or mentioned the need for key performance indicators (KPIs). Early-phase helix innovation ecosystems are mainly supported by qualitative or non-empirical data. This review explores multi-actor needs along with describing quantifiable KPIs at micro (end-user), meso (innovation hub) and macro (regional, national and international) levels. This integrated Penta hub approach will help to effectively plan, co-create, manage, analyse and utilize voluminous data, for example there are ca. 60,000 and 56,000 publications per year on artificial intelligence (AI) and medical devices respectively along, with some 35,000 adverse reports on devices submitted to the US FDA. This review addresses sustaining and potentially disruptive opportunities for decontamination and sterilization that includes the use of AI-enabled devices, bespoke training and sustainability.

Compared to cultured 2D cell monolayers, 3D multicellular spheroids are more realistic tumor models. Nonetheless, spheroids remain under-utilized in preclinical research, in part, because there is a lack of fluorescence sensors that can noninvasively interrogate all the individual cells within a spheroid. This present study describes a deep-red fluorogenic molecular probe for microscopic imaging of cells that contain a high level of nitroreductase enzyme activity as a biomarker of cell hypoxia. A first-generation version of the probe produced "turn-on" fluorescence in a 2D cell monolayer under hypoxic conditions; however, it was not useful in a 3D multicellular tumor spheroid because it only accumulated in the peripheral cells. To guide the probe structural optimization process, an intuitive theoretical membrane partition model was conceived to predict how a dosed probe will distribute within a 3D spheroid. The model identifies three limiting molecular diffusion pathways that are determined by a probe's membrane partition properties. A lipophilic probe with high membrane affinity rapidly becomes trapped in the membranes of the peripheral cells. In contrast, a very hydrophilic probe molecule with negligible membrane affinity diffuses rapidly through the spheroid intercellular space and rarely enters the cells. However, a probe molecule with intermediate membrane affinity undergoes sequential diffusion in and out of cells and distributes to all the cells within a spheroid. Using the model as a predictive tool, a second-generation fluorescent probe was prepared with a smaller and more hydrophilic molecular structure, and optical sectioning using structured illumination or light sheet microscopy revealed roughly even probe diffusion throughout a tumor spheroid. The membrane permeation model is likely to be broadly applicable for the structural optimization of various classes of molecules and nanoparticles to enable even distribution within a tumor spheroid.

Hemophilia A is a rare genetic condition that predominantly affects men and is characterized by a deficiency in Factor VIII clotting (FVIII). This research focuses on the development of a classification model to predict the severity of Hemophilia A, using data from point mutations in the FVIII protein. The study employs a variety of classification models, including RandomForest, XGBoost, and LightGBM, and performs a robust analysis of the data to select the most relevant features. The final model achieved an accuracy of 65.5 %, demonstrating significant performance against a simple gaussian naive bayes model that achieves 51.1 % of accuracy. Although the model cannot yet replace the FVIII measurement test in the blood for diagnostic purposes, the results represent a significant advance in Hemophilia A research. This work provides data analysis that deepens the understanding of the characteristics of the FVIII protein and contributes to the development of models capable of classifying the severity of this condition into its three possible classes: mild, moderate, or severe.

Gastrointestinal (GI) cancers are highly prevalent and considered a major global health challenge. Their approach has undergone a remarkable transformation over the past years due to the development of new technologies that enabled better outcomes regarding their diagnosis and management. These include artificial intelligence, robotics, next-generation sequencing and personalized medicine. Nonetheless, the integration of these advances into everyday clinical practice remains complex and challenging as we are still trying to figure out if these innovations tangibly improve oncological outcomes or if the current state of art should remain as the gold standard for the treatment of these patients. Additionally, there are also some issues regarding ethical subjects, data privacy, finances and governance. Precision surgery concept has evolved considerably over the past decades, especially for oncological patients. It aims to customize medical treatments and to operate on those patients who most likely will benefit from a specific surgical procedure. In the future, to improve GI oncological outcomes, a delicate balance between technological advances adoption and evidence-based care should be chased. As we move forward, the question will be to harness the power of innovation while keeping up the highest standards of patient care.

The multifaceted issue of childhood stunting in low- and middle-income countries has a profound and enduring impact on children's well-being, cognitive development, and future earning potential. Childhood stunting arises from a complex interplay of genetic, environmental, and socio-cultural factors. It requires a comprehensive approach across nutrition, education, healthcare, and poverty reduction sectors to mitigate its prevalence and short- and long-term effects. The Mexican case presents a distinct challenge, as the country has experienced the recent dissolution of social health security programmes, rising poverty rates, and reduced government expenditures for childhood well-being.

We propose a machine learning approach to understand the contribution of social and economic determinants to childhood stunting risk in Mexico. Using data from the 2006-2018 population-based Mexican National Health and Nutrition Surveys, six different machine learning classification algorithms were used to model and identify the most important predictors of childhood stunting.

Among the six classification algorithms tested, Extreme Gradient Boosting (XGB) obtained the highest Youden Index value, effectively balancing the correct classification of children with and without stunting. In the XGB model, the most important predictor for classifying childhood stunting is the household's socioeconomic status, followed by the state of residence, the child's age, indigenous population status, the household's portion of children under five years old, and the local area's deprivation level.

This paper contributes to understanding the structural determinants of stunting in children, emphasising the importance of implementing tailored interventions and policies, especially given our findings that highlight indigenous status and local deprivation as key predictors. In the context of diminishing health initiatives, this underscores the urgent need for specific, targeted, and sustainable actions to prevent and address a potential rise in stunting in similar settings.

social and economic deprivation, stunting, children, machine learning, XGB model, Mexico.

Maternal mortality refers to a mother's death owing to complications arising from childbirth or pregnancy. This issue is a forefront public health challenge around the globe which is pronounced in low- and middle-income countries, particularly in the sub-Saharan African regions where the burdens remain significantly high. Moreover, this problem is further complicated in developing countries due to limited access to antenatal care and the shortage of skilled birth attendants. So far, considerable improvements in the health status of many populations have been reported in developing countries. Nonetheless, the MDGs to reduce maternal and newborn mortality unmet in many SSA nations. Leveraging machine learning approaches allows us to better understand these constraints and predict skilled birth attendance among reproductive age women, providing actionable insights for policy and intervention.

This study aimed to predict skill birth attendance and identify its determinants among reproductive age women in 27 SSA countries using machine learning algorithm.

Using data from the Demographic and Health Surveys (2016-2024) across 27 SSA countries, we analyzed responses from 198,707 reproductive age women. The Random Forest classifier, complemented by SHAP for feature interpretability, was employed for prediction and analysis. Data preprocessing included K-nearest neighbor imputation for missing values, SMOTE for handling class imbalance, and Recursive Feature Elimination for feature selection. Model performance was evaluated using metrics such as accuracy, recall, F1 score, and AUC-ROC.

The Random Forest model demonstrated robust performance, achieving an AUC-ROC of 92%, recall of 96%, accuracy of 92%, precision of 93 and F1 score of 93%. The SHAP analysis identifies key predictors of skilled birth attendance, including facility delivery, maternal education, higher wealth index, urban residence, reduced distance to healthcare facilities, media exposure, and internet use.

The findings highlight the potential of machine learning to identify critical predictors of skilled birth attendance to inform targeted interventions. Addressing socioeconomic and educational disparities, enhancing healthcare access, and implementing tailored cessation programs are crucial to enhance skilled birth attendance in this vulnerable population.

Foundation models are pretrained on massive datasets. However, collecting medical datasets is expensive and time-consuming, and raises privacy concerns. Here we show that synthetic data generated via conditioning with disease labels can be leveraged for building high-performing medical foundation models. We pretrained a retinal foundation model, first with approximately one million synthetic retinal images with physiological structures and feature distribution consistent with real counterparts, and then with only 16.7% of the 904,170 real-world colour fundus photography images required in a recently reported retinal foundation model (RETFound). The data-efficient model performed as well or better than RETFound across nine public datasets and four diagnostic tasks; and for diabetic-retinopathy grading, it used only 40% of the expert-annotated training data used by RETFound. We also support the generalizability of the data-efficient strategy by building a classifier for the detection of tuberculosis on chest X-ray images. The text-conditioned generation of synthetic data may enhance the performance and generalization of medical foundation models.

In recent decades, 3D bioprinting has garnered significant research attention due to its ability to manipulate biomaterials and cells to create complex structures precisely. However, due to technological and cost constraints, the clinical translation of 3D bioprinted products (BPPs) from bench to bedside has been hindered by challenges in terms of personalization of design and scaling up of production. Recently, the emerging applications of artificial intelligence (AI) technologies have significantly improved the performance of 3D bioprinting. However, the existing literature remains deficient in a methodological exploration of AI technologies' potential to overcome these challenges in advancing 3D bioprinting toward clinical application. This paper aims to present a systematic methodology for AI-driven 3D bioprinting, structured within the theoretical framework of Quality by Design (QbD). This paper commences by introducing the QbD theory into 3D bioprinting, followed by summarizing the technology roadmap of AI integration in 3D bioprinting, including multi-scale and multi-modal sensing, data-driven design, and in-line process control. This paper further describes specific AI applications in 3D bioprinting's key elements, including bioink formulation, model structure, printing process, and function regulation. Finally, the paper discusses current prospects and challenges associated with AI technologies to further advance the clinical translation of 3D bioprinting.

Machine learning models for the diagnosis of breast cancer can facilitate the prediction of cancer risk and subsequent patient management among other clinical tasks. For the models to impact clinical practice, they ought to follow standard workflows, help interpret mammography and ultrasound data, evaluate clinical contextual information, handle incomplete data and be validated in prospective settings. Here we report the development and testing of a multimodal model leveraging mammography and ultrasound modules for the stratification of breast cancer risk based on clinical metadata, mammography and trimodal ultrasound (19,360 images of 5,216 breasts) from 5,025 patients with surgically confirmed pathology across medical centres and scanner manufacturers. Compared with the performance of experienced radiologists, the model performed similarly at classifying tumours as benign or malignant and was superior at pathology-level differential diagnosis. With a prospectively collected dataset of 191 breasts from 187 patients, the overall accuracies of the multimodal model and of preliminary pathologist-level assessments of biopsied breast specimens were similar (90.1% vs 92.7%, respectively). Multimodal models may assist diagnosis in oncology.

Medical research within areas of deep learning, particularly in computer vision for medical imaging, has shown promise over the past decade with an increasing volume of technical papers published in orthopaedics related to imaging artificial intelligence (AI). However, as more tools and models are developed and deployed, it is easy for clinicians to get overwhelmed with the different types of models, leaving "artificial intelligence" as an empty buzzword where true value can be unclear. As with surgery, the techniques of deep learning require thoughtful application and cannot follow a one-size-fits-all approach as different problems require differential levels of technical complexity with model application. Moreover, the application of AI-based clinical tools should be both adjunctive and transparent in their stepwise integration within clinical medicine to provide additive insight. As a medical profession, we must together decide how, when, and where we want AI-based applications to offer insight.

This systematic review investigates the use of machine learning (ML) algorithms in predicting survival outcomes for ovarian cancer (OC) patients. Key prognostic endpoints, including overall survival (OS), recurrence-free survival (RFS), progression-free survival (PFS), and treatment response prediction (TRP), are examined to evaluate the effectiveness of these algorithms and identify significant features that influence predictive accuracy.

A thorough search of four major databases-PubMed, Scopus, Web of Science, and Cochrane-resulted in 2400 articles published within the last decade, with 32 studies meeting the inclusion criteria. Notably, most publications emerged after 2021. Commonly used algorithms for survival prediction included random forest, support vector machines, logistic regression, XGBoost, and various deep learning models. Evaluation metrics such as area under the curve (AUC) (18 studies), concordance index (C-index) (11 studies), and accuracy (11 studies) were frequently employed. Age at diagnosis, tumor stage, CA-125 levels, and treatment-related factors were consistently highlighted as significant predictors, emphasizing their relevance in OC prognosis.

ML models demonstrate considerable potential for predicting OC survival outcomes; however, challenges persist regarding model accuracy and interpretability. Incorporating diverse data types-such as clinical, imaging, and molecular datasets-holds promise for enhancing predictive capabilities. Future advancements will depend on integrating heterogeneous data sources with multimodal ML approaches, which are crucial for improving prognostic precision in OC.

This scoping review systematically maps externally validated machine learning (ML)-based models in cancer patient care, quantifying their performance, and clinical utility, and examining relationships between models, cancer types, and clinical decisions. By synthesizing evidence, this study identifies, strengths, limitations, and areas requiring further research.

The review followed the Joanna Briggs Institute's methodology, Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines, and the Population, Concept, and Context mnemonic. Searches were conducted across Embase, IEEE Xplore, PubMed, Scopus, and Web of Science (January 2014-September 2022), targeting English-language quantitative studies in Q1 journals (SciMago Journal and Country Ranking > 1) that used ML to evaluate clinical outcomes for human cancer patients with commonly available data. Eligible models required external validation, clinical utility assessment, and performance metric reporting. Studies involving genetics, synthetic patients, plants, or animals were excluded. Results were presented in tabular, graphical, and descriptive form.

From 4023 deduplicated abstracts and 636 full-text reviews, 56 studies (2018-2022) met the inclusion criteria, covering diverse cancer types and applications. Convolutional neural networks were most prevalent, demonstrating high performance, followed by gradient- and decision tree-based algorithms. Other algorithms, though underrepresented, showed promise. Lung and digestive system cancers were most frequently studied, focusing on diagnosis and outcome predictions. Most studies were retrospective and multi-institutional, primarily using image-based data, followed by text-based and hybrid approaches. Clinical utility assessments involved 499 clinicians and 12 tools, indicating improved clinician performance with AI assistance and superior performance to standard clinical systems.

Interest in ML-based clinical decision-making has grown in recent years alongside increased multi-institutional collaboration. However, small sample sizes likely impacted data quality and generalizability. Persistent challenges include limited international validation across ethnicities, inconsistent data sharing, disparities in validation metrics, and insufficient calibration reporting, hindering model comparison reliability.

Successful integration of ML in oncology decision-making requires standardized data and methodologies, larger sample sizes, greater transparency, and robust validation and clinical utility assessments.

Financed by FCT-Fundação para a Ciência e a Tecnologia (Portugal, project LA/P/0063/2020, grant 2021.09040.BD) as part of CSS's Ph.D. This work was not registered.

The Epic End of Life Care Index (EOLCI) predicts 1-year mortality for a general adult population using medical record data. It is deployed at various medical centers, but we are not aware of an independent validation. We evaluated its performance for predicting 1-year mortality in patients with metastatic cancer, comparing it against an academic machine learning model designed for cancer patients. We focused on this patient population because of their high short-term mortality risk and because we had access to the comparator model predictions.

This retrospective analysis included adult outpatients with metastatic cancer from four outpatient sites. Performance metrics included AUC for 1-year mortality and positive predictive value of high-risk score.

There were 1,399 patients included. Median age at first EOLCI prediction was 67 and 55% were female. A total of 1,283 patients were evaluable for 1-year mortality; of these, 297 (23%) died within 1 year. AUC for 1-year mortality for EOLCI and academic model was 0.73 (95% CI [0.70-0.76]) and 0.82 (95% CI [0.80-0.85]), respectively. Positive predictive value was 0.38 and 0.65, respectively.

The EOLCI's discrimination performance was lower than the vendor-stated value (AUC of 0.86) and the academic model's performance. Vendor-supplied machine learning models should be independently validated, particularly in specialized patient populations, to ensure accuracy and reliability.

In recent decades, the integration of artificial intelligence (AI) into health care has revolutionized diagnostics, treatment customization, and delivery. In low-resource settings, AI offers significant potential to address health care disparities exacerbated by shortages of medical professionals and other resources. However, implementing AI effectively and responsibly in these settings requires careful consideration of context-specific needs and barriers to equitable care. This article explores the practical deployment of AI in low-resource environments through a review of existing literature and interviews with experts, ranging from health care providers and administrators to AI tool developers and government consultants. The authors highlight 4 critical areas for effective AI deployment: infrastructure requirements, deployment and data management, education and training, and responsible AI practices. By addressing these aspects, the proposed framework aims to guide sustainable AI integration, minimizing risk, and enhancing health care access in underserved regions.

Wound management requires the measurement of the wound parameters such as its shape and area. However, computerized analysis of the wound suffers the challenge of inexact segmentation of the wound images due to limited or inaccurate labels. It is a common scenario that the source domain provides an abundance of labeled data, while the target domain provides only limited labels. To overcome this, we propose a novel approach that combines self-training learning and mixup augmentation. The neural network is trained on the source domain to generate weak labels on the target domain via the self-training process. In the second stage, generated labels are mixed up with labels from the source domain to retrain the neural network and enhance generalization across diverse datasets. The efficacy of our approach was evaluated using the DFUC 2022, FUSeg, and RMIT datasets, demonstrating substantial improvements in segmentation accuracy and robustness across different data distributions. Specifically, in single-domain experiments, segmentation on the DFUC 2022 dataset scored a dice score of 0.711, while the score on the FUSeg dataset achieved 0.859. For domain adaptation, when these datasets were used as target datasets, the dice scores were 0.714 for DFUC 2022 and 0.561 for FUSeg.

Computed tomography (CT) of the head and neck is crucial for diagnosing internal structures. The demand for substituting traditional CT with cone beam CT (CBCT) exists because of its cost-effectiveness and reduced radiation exposure. However, CBCT cannot accurately depict airway shapes owing to image noise. This study proposes a strategy utilizing a cycle-consistent generative adversarial network (cycleGAN) for denoising CBCT images with various loss functions and augmentation strategies, resulting in the generation of denoised synthetic CT (sCT) images. Furthermore, through a rule-based approach, we were able to automatically segment the upper airway in sCT images with high accuracy. Additionally, we conducted an analysis of the impact of finely segmented nasal cavities on airflow using computational fluid dynamics (CFD).

We trained the cycleGAN model using various loss functions and compared the quality of the sCT images generated by each model. We improved the artifact removal performance by incorporating CT images with added Gaussian noise augmentation into the training dataset. We developed a rule-based automatic segmentation methodology using threshold and watershed algorithms to compare the accuracy of airway segmentation for noise-reduced sCT and original CBCT. Furthermore, we validated the significance of the nasal cavity by conducting CFD based on automatically segmented shapes obtained from sCT.

The generated sCT images exhibited improved quality, with the mean absolute error decreasing from 161.60 to 100.54, peak signal-to-noise ratio increasing from 22.33 to 28.65, and structural similarity index map increasing from 0.617 to 0.865. Furthermore, by comparing the airway segmentation performances of CBCT and sCT using our proposed automatic rule-based algorithm, the Dice score improved from 0.849 to 0.960. Airway segmentation performance is closely associated with the accuracy of fluid dynamics simulations. Detailed airway segmentation is crucial for altering flow dynamics and contributes significantly to diagnostics.

Our deep learning methodology enhances the image quality of CBCT to provide anatomical information to medical professionals and enables precise and accurate biomechanical analysis. This allows clinicians to obtain precise quantitative metrics and facilitates accurate assessment.

Postoperative fever (POF) is a common occurrence in patients undergoing major surgery, presenting challenges and burdens for both patients and surgeons yet. This study endeavors to examine the incidence, identify risk factors, and establish a machine learning-based predictive model for POF following surgery of oral cancer.

A total of seven hundred and twenty-seven consecutive patients undergoing radical resection of oral cancer were retrospectively investigated. The analysis encompassed 34 parameters, incorporating demographic and clinical characteristics, biochemical and hematological assay results, surgical-related data, hospitalization costs and stay in hospital. Six machine learning models were compared by the area under the receiver operating characteristic curve (AUC). The best-performing models were selected for further analyze, including feature importance evaluation and nomogram analysis, identifying key POF risk factors, and establish a comprehensive prediction model.

A total of 466 patients with surgery for oral cancer met the criteria, with an average age of (54.2 ± 11.1) years, including an POF group (n = 197) and a non-POF group (n = 269). The fever group has greater hospitalization costs, longer lengths of stay, and higher infection biochemical indicators (leucocyte ratio and erythrocyte sedimentation rate). Furthermore, Among the 6 machine learning models, logistic regression models performed best, with the higher AUC and accuracy. In univariate and multivariate logistic analysis showed that age, sex, reoperation, Charlson Comorbidity Index score (CCI), leukocyte, bleeding and blood transfusion were independent risk factors for POF of patients following surgery in oral cancer. Then seven variables were selected to establish the nomogram for predict the probability of POF by nomogram algorithm.

Postoperative fever patients following radical resection of oral cancer have greater burden. Machine learning algorithms can be effectively used to identify potential risk factors of POF, which may enhance individualized treatment plans in oral cancer patient during perioperative period.

Heart failure (HF) is a significant global health problem, affecting approximately 64.34 million people worldwide. The worsening of HF, also known as HF decompensation, is a major factor behind hospitalizations, contributing to substantial health care costs related to this condition.

This study aimed to explore the perspectives of health care professionals and data scientists regarding the relevance, challenges, and potential benefits of using machine learning (ML) models to predict decompensation from patients with HF.

A total of 13 individual, semistructured, qualitative interviews were conducted in Portugal between October 31, 2022, and June 23, 2023. Participants represented different health care specialties and were selected from different contexts and regions of the country to ensure a comprehensive understanding of the topic. Data saturation was determined as the point at which no new themes emerged from participants' perspectives, ensuring a sufficient sample size for analysis. The interviews were audio recorded, transcribed, and analyzed using MAXQDA (VERBI Software GmbH) through a reflexive thematic analysis. Two researchers (JS and AH) coded the interviews to ensure the consistency of the codes. Ethical approval was granted by the NOVA National School of Public Health ethics committee (CEENSP 14/2022), and informed consent was obtained from all participants.

The participants recognized the potential benefits of ML models for early detection, risk stratification, and personalized care of patients with HF. The importance of selecting appropriate variables for model development, such as rapid weight gain and symptoms, was emphasized. The use of wearables for recording vital signs was considered necessary, although challenges related to adoption among older patients were identified. Risk stratification emerged as a crucial aspect, with the model needing to identify patients at high-, medium-, and low-risk levels. Participants emphasized the need for a response model involving health care professionals to validate ML-generated alerts and determine appropriate interventions.

The study's findings highlight ML models' potential benefits and challenges for predicting HF decompensation. The relevance of ML models for improving patient outcomes, reducing health care costs, and promoting patient engagement in disease management is highlighted. Adequate variable selection, risk stratification, and response models were identified as essential components for the effective implementation of ML models in health care. In addition, the study identified technical, regulatory and ethical, and adoption and acceptance challenges that need to be overcome for the successful integration of ML models into clinical workflows. Interpretation of the findings suggests that future research should focus on more extensive and diverse samples, incorporate the patient perspective, and explore the impact of ML models on patient outcomes and personalized care in HF management. Incorporation of this study's findings into practice is expected to contribute to developing and implementing ML-based predictive models that positively impact HF management.

Diffusion models, variational autoencoders, and generative adversarial networks (GANs) are three common types of generative artificial intelligence models for image generation. Among these, GANs are the most frequently used for medical image generation and are often employed for data augmentation in various studies. However, due to the adversarial nature of GANs, where the generator and discriminator compete against each other, the training process can sometimes end with the model unable to generate meaningful images or even producing noise. This phenomenon is rarely discussed in the literature, and no studies have proposed solutions to address this issue. Such outcomes can introduce significant bias when GANs are used for data augmentation in medical image training. Moreover, GANs often require substantial computational power and storage, adding to the challenges. In this study, we used deep convolutional GANs for chest X-ray generation, and three typical training outcomes were found. Two scenarios generated meaningful medical images and one failed to produce usable images. By analyzing the loss history during training, we observed that the regression line of the overall losses tends to diverge slowly. After excluding outlier losses, we found that the slope of the regression line within the stable loss segment indicates the optimal point to terminate training, ensuring the generation of meaningful medical images.

Multi-modality imaging is widely used in clinical practice and biomedical research to gain a comprehensive understanding of an imaging subject. Currently, multi-modality imaging is accomplished by post hoc fusion of independently reconstructed images under the guidance of mutual information or spatially registered hardware, which limits the accuracy and utility of multi-modality imaging. Here, we investigate a data-driven multi-modality imaging (DMI) strategy for synergetic imaging of CT and MRI. We reveal two distinct types of features in multi-modality imaging, namely intra- and inter-modality features, and present a multi-sensor learning (MSL) framework to utilize the crossover inter-modality features for augmented multi-modality imaging. The MSL imaging approach breaks down the boundaries of traditional imaging modalities and allows for optimal hybridization of CT and MRI, which maximizes the use of sensory data. We showcase the effectiveness of our DMI strategy through synergetic CT-MRI brain imaging. The principle of DMI is quite general and holds enormous potential for various DMI applications across disciplines.

Health event prediction is empowered by the rapid and wide application of electronic health records (EHR). In the Intensive Care Unit (ICU), precisely predicting the health related events in advance is essential for providing treatment and intervention to improve the patients outcomes. EHR is a kind of multi-modal data containing clinical text, time series, structured data, etc. Most health event prediction works focus on a single modality, e.g., text or tabular EHR. How to effectively learn from the multi-modal EHR for health event prediction remains a challenge. Inspired by the strong capability in text processing of large language model (LLM), we propose the framework CKLE for health event prediction by distilling the knowledge from LLM and learning from multi-modal EHR. There are two challenges of applying LLM in the health event prediction, the first one is most LLM can only handle text data rather than other modalities, e.g., structured data. The second challenge is the privacy issue of health applications requires the LLM to be locally deployed, which may be limited by the computational resource. CKLE solves the challenges of LLM scalability and portability in the healthcare domain by distilling the cross-modality knowledge from LLM into the health event predictive model. To fully take advantage of the strong power of LLM, the raw clinical text is refined and augmented with prompt learning. The embedding of clinical text are generated by LLM. To effectively distill the knowledge of LLM into the predictive model, we design a cross-modality knowledge distillation (KD) method. A specially designed training objective will be used for the KD process with the consideration of multiple modality and patient similarity. The KD loss function consists of two parts. The first one is cross-modality contrastive loss function, which models the correlation of different modalities from the same patient. The second one is patient similarity learning loss function to model the correlations between similar patients. The cross-modality knowledge distillation can distill the rich information in clinical text and the knowledge of LLM into the predictive model on structured EHR data. To demonstrate the effectiveness of CKLE, we evaluate CKLE on two health event prediction tasks in the field of cardiology, heart failure prediction and hypertension prediction. We select the 7125 patients from MIMIC-III dataset and split them into train/validation/test sets. We can achieve a maximum 4.48% improvement in accuracy compared to state-of-the-art predictive model designed for health event prediction. The results demonstrate CKLE can surpass the baseline prediction models significantly on both normal and limited label settings. We also conduct the case study on cardiology disease analysis in the heart failure and hypertension prediction. Through the feature importance calculation, we analyse the salient features related to the cardiology disease which corresponds to the medical domain knowledge. The superior performance and interpretability of CKLE pave a promising way to leverage the power and knowledge of LLM in the health event prediction in real-world clinical settings.

Prehospital medical care is a major challenge for both civilian and military situations as resources are limited, yet critical triage and treatment decisions must be rapidly made. Prehospital medicine is further complicated during mass casualty situations or remote applications that require more extensive medical treatments to be monitored. It is anticipated on the future battlefield where air superiority will be contested that prolonged field care will extend to as much 72 h in a prehospital environment. Traditional medical monitoring is not practical in these situations and, as such, wearable sensor technology may help support prehospital medicine. However, sensors alone are not sufficient in the prehospital setting where limited personnel without specialized medical training must make critical decisions based on physiological signals. Machine learning-based clinical decision support systems can instead be utilized to interpret these signals for diagnosing injuries, making triage decisions, or driving treatments. Here, we summarize the challenges of the prehospital medical setting and review wearable sensor technology suitability for this environment, including their use with medical decision support triage or treatment guidance options. Further, we discuss recommendations for wearable healthcare device development and medical decision support technology to better support the prehospital medical setting. With further design improvement and integration with decision support tools, wearable healthcare devices have the potential to simplify and improve medical care in the challenging prehospital environment.

Background/Objectives: This study aimed to develop and validate a machine learning (ML)-based metabolic syndrome (MetS) risk prediction model. Methods: We examined data from 6155 participants of the China Health and Retirement Longitudinal Study (CHARLS) in 2011. The LASSO regression feature selection identified the best MetS predictors. Nine ML-based algorithms were adopted to build predictive models. The model performance was validated using cohort data from the Korea National Health and Nutrition Examination Survey (KNHANES) (n = 5297), the United Kingdom (UK) Biobank (n = 218,781), and the National Health and Nutrition Examination Survey (NHANES) (n = 2549). Results: The multilayer perceptron (MLP)-based model performed best in the CHARLS cohort (AUC = 0.8908; PRAUC = 0.8073), the logistic model in the KNHANES cohort (AUC = 0.9101, PRAUC = 0.8116), the xgboost model in the UK Biobank cohort (AUC = 0.8556, PRAUC = 0.6246), and the MLP model in the NHANES cohort (AUC = 0.9055, PRAUC = 0.8264). Conclusions: Our MLP-based model has the potential to serve as a clinical application for detecting MetS in different populations.

Drug repurposing-identifying new therapeutic uses for approved drugs-is often a serendipitous and opportunistic endeavour to expand the use of drugs for new diseases. The clinical utility of drug-repurposing artificial intelligence (AI) models remains limited because these models focus narrowly on diseases for which some drugs already exist. Here we introduce TxGNN, a graph foundation model for zero-shot drug repurposing, identifying therapeutic candidates even for diseases with limited treatment options or no existing drugs. Trained on a medical knowledge graph, TxGNN uses a graph neural network and metric learning module to rank drugs as potential indications and contraindications for 17,080 diseases. When benchmarked against 8 methods, TxGNN improves prediction accuracy for indications by 49.2% and contraindications by 35.1% under stringent zero-shot evaluation. To facilitate model interpretation, TxGNN's Explainer module offers transparent insights into multi-hop medical knowledge paths that form TxGNN's predictive rationales. Human evaluation of TxGNN's Explainer showed that TxGNN's predictions and explanations perform encouragingly on multiple axes of performance beyond accuracy. Many of TxGNN's new predictions align well with off-label prescriptions that clinicians previously made in a large healthcare system. TxGNN's drug-repurposing predictions are accurate, consistent with off-label drug use, and can be investigated by human experts through multi-hop interpretable rationales.

This study describes the development and deployment of a machine learning (ML) model called Vent.io to predict mechanical ventilation (MV).

We trained Vent.io using electronic health record data of adult patients admitted to the intensive care units (ICUs) of the University of California San Diego (UCSD) Health System. We prospectively deployed Vent.io using a real-time platform at UCSD and evaluated the performance of Vent.io for a 1-month period in silent mode and on the MIMIC-IV dataset. As part of deployment, we included a Predetermined Changed Control Plan (PCCP) for continuous model monitoring that triggers model fine-tuning if performance drops below a specified area under the receiver operating curve (AUC) threshold of 0.85.

The Vent.io model had a median AUC of 0.897 (IQR: 0.892-0.904) with specificity of 0.81 (IQR: 0.812-0.841) and positive predictive value (PPV) of 0.174 (IQR: 0.148-0.176) at a fixed sensitivity of 0.6 during 10-fold cross validation and an AUC of 0.908, sensitivity of 0.632, specificity of 0.849, and PPV of 0.235 during prospective deployment. Vent.io had an AUC of 0.73 on the MIMIC-IV dataset, triggering model fine-tuning per the PCCP as the AUC was below the minimum of 0.85. The fine-tuned Vent.io model achieved an AUC of 0.873.

Deterioration of model performance is a significant challenge when deploying ML models prospectively or at different sites. Implementation of a PCCP can help models adapt to new patterns in data and maintain generalizability.

Vent.io is a generalizable ML model that has the potential to improve patient care and resource allocation for ICU patients with need for MV.

Digital twins represent a key technology for precision health. Medical digital twins consist of computational models that represent the health state of individual patients over time, enabling optimal therapeutics and forecasting patient prognosis. Many health conditions involve the immune system, so it is crucial to include its key features when designing medical digital twins. The immune response is complex and varies across diseases and patients, and its modelling requires the collective expertise of the clinical, immunology, and computational modelling communities. This review outlines the initial progress on immune digital twins and the various initiatives to facilitate communication between interdisciplinary communities. We also outline the crucial aspects of an immune digital twin design and the prerequisites for its implementation in the clinic. We propose some initial use cases that could serve as "proof of concept" regarding the utility of immune digital technology, focusing on diseases with a very different immune response across spatial and temporal scales (minutes, days, months, years). Lastly, we discuss the use of digital twins in drug discovery and point out emerging challenges that the scientific community needs to collectively overcome to make immune digital twins a reality.

Biomedical image analysis plays a critical role in the healthcare system [...].

Routine monitoring of renal and hepatic function during chemotherapy ensures that treatment-related organ damage has not occurred and clearance of subsequent treatment is not hindered; however, frequency and timing are not optimal. Model bias and data heterogeneity concerns have hampered the ability of machine learning (ML) to be deployed into clinical practice. This study aims to develop models that could support individualised decisions on the timing of renal and hepatic monitoring while exploring the effect of data shift on model performance.

We used retrospective data from three UK hospitals to develop and validate ML models predicting unacceptable rises in creatinine/bilirubin post cycle 3 for patients undergoing treatment for the following cancers: breast, colorectal, lung, ovarian and diffuse large B-cell lymphoma.

We extracted 3614 patients with no missing blood test data across cycles 1-6 of chemotherapy treatment. We improved on previous work by including predictions post cycle 3. Optimised for sensitivity, we achieve F2 scores of 0.7773 (bilirubin) and 0.6893 (creatinine) on unseen data. Performance is consistent on tumour types unseen during training (F2 bilirubin: 0.7423, F2 creatinine: 0.6820).

Our technique highlights the effectiveness of ML in clinical settings, demonstrating the potential to improve the delivery of care. Notably, our ML models can generalise to unseen tumour types. We propose gold-standard bias mitigation steps for ML models: evaluation on multisite data, thorough patient population analysis, and both formalised bias measures and model performance comparisons on patient subgroups. We demonstrate that data aggregation techniques have unintended consequences on model bias.

In the realm of intelligent healthcare, there is a growing ambition to reshape medical services through the integration of artificial intelligence (AI). However, conventional machine learning faces inherent challenges such as privacy issues, delayed updates, and protracted training times, particularly due to the hesitance of medical institutions to directly share sensitive data, with possible noises. In response to these concerns, a Quantum-Assisted Federated Intelligent Diagnosis Algorithm ( β -QuAFIDA) is proposed, applied into real medical data. Leveraging the capabilities of the 5G mobile network, this approach works the connection between Internet of Medical Things (IoMT) devices through the 5G, synchronizing training and updating the server model without disrupting their real-world applications. In our quest to safeguard patient data and enhance training efficiency, our study employs an innovative heuristic approach marked by a nested loop structure. Specifically, the inner loop is dedicated to training the beta-variational quantum eigensolver ( β -VQE) to approximate the expectation values of the proposed algorithm; the outer loop trains the β -QuAFIDA to reduce the relative entropy towards the target. This approach involves a balance between privacy considerations and the urgency of training. Results demonstrate that representations with low-rank attained through β -QuAFIDA offer an effective approach for acquiring low-rank states. This research signifies a step forward in the synergy between AI and 5G technologies, presenting a novel avenue for the advancement of intelligent healthcare.

The retinal age gap (RAG) is emerging as a potential biomarker for various diseases of the human body, yet its utility depends on machine learning models capable of accurately predicting biological retinal age from fundus images. However, training generalizable models is hindered by potential shortages of diverse training data. To overcome these obstacles, this work develops a novel and computationally efficient distributed learning framework for retinal age prediction.

The proposed framework employs a memory-efficient 8-bit quantized version of RETFound, a cutting-edge foundation model for retinal image analysis, to extract features from fundus images. These features are then used to train an efficient linear regression head model for predicting retinal age. The framework explores federated learning (FL) as well as traveling model (TM) approaches for distributed training of the linear regression head. To evaluate this framework, we simulate a client network using fundus image data from the UK Biobank. Additionally, data from patients with type 1 diabetes from the UK Biobank and the Brazilian Multilabel Ophthalmological Dataset (BRSET) were utilized to explore the clinical utility of the developed methods.

Our findings reveal that the developed distributed learning framework achieves retinal age prediction performance on par with centralized methods, with FL and TM providing similar performance (mean absolute error of 3.57 ± 0.18 years for centralized learning, 3.60 ± 0.16 years for TM, and 3.63 ± 0.19 years for FL). Notably, the TM was found to converge with fewer local updates than FL. Moreover, patients with type 1 diabetes exhibited significantly higher RAG values than healthy controls in all models, for both the UK Biobank and BRSET datasets (P < .001).

The high computational and memory efficiency of the developed distributed learning framework makes it well suited for resource-constrained environments.

The capacity of this framework to integrate data from underrepresented populations for training of retinal age prediction models could significantly enhance the accessibility of the RAG as an important disease biomarker.

Hemorrhage leading to life-threatening shock is a common and critical problem in both civilian and military medicine. Due to complex physiological compensatory mechanisms, traditional vital signs may fail to detect patients' impending hemorrhagic shock in a timely manner when life-saving interventions are still viable. To address this shortcoming of traditional vital signs in detecting hemorrhagic shock, we have attempted to identify metrics that can predict blood loss. We have previously combined feature extraction and machine learning methodologies applied to arterial waveform analysis to develop advanced metrics that have enabled the early and accurate detection of impending shock in a canine model of hemorrhage, including metrics that estimate blood loss such as the Blood Loss Volume Metric, the Percent Estimated Blood Loss metric, and the Hemorrhage Area metric. Importantly, these metrics were able to identify impending shock well before traditional vital signs, such as blood pressure, were altered enough to identify shock. Here, we apply these advanced metrics developed using data from a canine model to data collected from a swine model of controlled hemorrhage as an interim step towards showing their relevance to human medicine. Based on the performance of these advanced metrics, we conclude that the framework for developing these metrics in the previous canine model remains applicable when applied to a swine model and results in accurate performance in these advanced metrics. The success of these advanced metrics in swine, which share physiological similarities to humans, shows promise in developing advanced blood loss metrics for humans, which would result in increased positive casualty outcomes due to hemorrhage in civilian and military medicine.

This study aims to use machine learning model to predict laboratory aspirin resistance (AR) in Chinese stroke patients by incorporating patient characteristics and single nucleotide polymorphisms of GP1BA and LTC4S. 2405 patients were analyzed to measure the Mutation frequency of GP1BA rs6065 and LTC4S rs730012. 112 patients with first-stroke arteriostenosis were prospectively enrolled to establish machine learning model. GP1BA rs6065 mutation frequency is 5.26% and LTC4S rs730012 is 14.78%. GP1BA rs6065 CT patients have more sensitivity to aspirin than CC genotype. Simple linear regression identified significant associations with age, smoking, HDL and GP1BA rs6065. Random forest (RF) and extreme gradient boosting (XGBoost) demonstrated predictive capabilities for AR. Findings suggest pre-identifying GP1BA rs6065 could optimize aspirin treatment, enabling personalized care and future research avenues.

This study aims to examine the proportions of patients referred to mental health, social and voluntary, community and social enterprise (VCSE) services and general practice and to assess care gaps among people presenting to the hospital following self-harm.

Population-based observational study. Data were extracted from hospital records.

Three emergency departments (EDs) in Manchester, UK.

26 090 patients aged 15+ years who presented to participating EDs following self-harm and who received a psychosocial assessment by a mental health specialist.

Primary outcome measures are as follows: care gaps, estimated from the proportion of patients with evidence of social and mental health needs with no new or active referral to mental health, social and VCSE services. Secondary outcome measures are as follows: proportions of referrals by groups of patients, estimated mental health and social needs of patients. Indicators of mental health and social need were developed with academic clinicians (psychiatrist, general practitioner and social worker) and expert lived experience contributors.

96.2% (25 893/26 909) of individuals were estimated as having mental health needs. Among this group, 29.9% (6503/21 719) had no new or active referral to mental health services (indicating a care gap). Mental healthcare gaps were greater in men and those who were aged under 35 years, from a black, South Asian or Chinese ethnic group, living in the most deprived areas and had no mental health diagnosis, or alcohol, substance misuse, anxiety or trauma-related disorder. 52.8% (14 219/26 909) had social needs, with care gaps greater for men, individuals aged 45-64 and those who were unemployed or had a diagnosed mental disorder.

Care gaps were higher among hospital-presenting groups known to have increased risks of suicide: men, those in middle age, unemployed individuals and those misusing substances. Improved access to mental health, social and VCSE services and general practice care is vital to reduce inequities in access to self-harm aftercare.

Cancer staging is an essential clinical attribute informing patient prognosis and clinical trial eligibility. However, it is not routinely recorded in structured electronic health records. Here, we present BB-TEN: Big Bird - TNM staging Extracted from Notes, a generalizable method for the automated classification of TNM stage directly from pathology report text. We train a BERT-based model using publicly available pathology reports across approximately 7000 patients and 23 cancer types. We explore the use of different model types, with differing input sizes, parameters, and model architectures. Our final model goes beyond term-extraction, inferring TNM stage from context when it is not included in the report text explicitly. As external validation, we test our model on almost 8000 pathology reports from Columbia University Medical Center, finding that our trained model achieved an AU-ROC of 0.815-0.942. This suggests that our model can be applied broadly to other institutions without additional institution-specific fine-tuning.

Prediction of short-term outcomes in young patients with acute ischemic stroke (AIS) may assist in making therapy decisions. Machine learning (ML) is increasingly used in healthcare due to its high accuracy. This study aims to use a ML-based predictive model for poor 3-month functional outcomes in young AIS patients and to compare the predictive performance of ML models with the logistic regression model.

We enrolled AIS patients aged between 18 and 50 years from the Third Chinese National Stroke Registry (CNSR-III), collected between 2015 and 2018. A modified Rankin Scale (mRS) ≥ 3 was a poor functional outcome at 3 months. Four ML tree models were developed: The extreme Gradient Boosting (XGBoost), Light Gradient Boosted Machine (lightGBM), Random Forest (RF), and The Gradient Boosting Decision Trees (GBDT), compared with logistic regression. We assess the model performance based on both discrimination and calibration.

A total of 2268 young patients with a mean age of 44.3 ± 5.5 years were included. Among them, (9%) had poor functional outcomes. The mRS at admission, living alone conditions, and high National Institutes of Health Stroke Scale (NIHSS) at discharge remained independent predictors of poor 3-month outcomes. The best AUC in the test group was XGBoost (AUC = 0.801), followed by GBDT, RF, and lightGBM (AUCs of 0.795, 0, 794, and 0.792, respectively). The XGBoost, RF, and lightGBM models were significantly better than logistic regression (P < 0.05).

ML outperformed logistic regression, where XGBoost the boost was the best model for predicting poor functional outcomes in young AIS patients. It is important to consider living alone conditions with high severity scores to improve stroke prognosis.

Measures in operations management are pivotal for monitoring and assessing various aspects of hospital performance. Existing literature highlights the importance of regularly updating key management measures to reflect changing trends and organizational goals. Advancements in machine learning (ML) have presented promising opportunities for enhancing the process of updating operations management measures. However, their specific application and performance remain relatively unexplored. We aimed to investigate the feasibility and effectiveness of using common ML techniques to identify and update key measures in hospital operations management.

Historical data on 43 measures on financial balance and quality of care under 4 categories were retrieved from the BI system of a regional health system in Central China. The dataset included 17 surgical and 15 non-surgical departments over 48 months. Four common ML techniques, linear models (LM), random forest (RF), partial least squares (PLS), and neural networks (NN), were used to identify the most important measures. Ordinary least square was employed to investigate the impact of the top 10 measures. A ground truth validation compared the ML-identified key measures against the humanly decided strategic measures from annual meeting minutes.

For financial balancing, inpatient treatment revenue was an important measure in 3/4 years, followed by equipment depreciation costs. The measures identified using the same technique differed between years, though RF and PLS yielded relatively consistent results. For quality of care, none of the ML-identified measures repeated over the years. Those consistently important over four years differed almost entirely among four techniques. On ground truth validation, the 2016-2019 ML-identified measures were among the humanly identified measures, with the exception of equipment depreciation from the 2019 dataset. All the ML-identified measures for quality of care failed to coincide with the humanly decided measures.

Using ML to identify key hospital operational measures is viable but performance of ML techniques vary considerably. RF performs best among the four techniques in identifying key measures in financial balance. None of the ML techniques seem effective for identifying quality of care measures. ML is suggested as a decision support tool to remind and inspire decision-makers in certain aspects of hospital operations management.

Precision and personalized medicine, the process by which patient management is tailored to individual circumstances, are now terms that are familiar to cardiologists, despite it still being an emerging field. Although precision medicine relies most often on the underlying biology and pathophysiology of a patient's condition, personalized medicine relies on digital biomarkers generated through algorithms. Given the complexity of the underlying data, these digital biomarkers are most often generated through machine-learning algorithms. There are a number of analytic considerations regarding the creation of digital biomarkers that are discussed in this review, including data preprocessing, time dependency and gating, dimensionality reduction, and novel methods, both in the realm of supervised and unsupervised machine learning. Some of these considerations, such as sample size requirements and measurements of model performance, are particularly challenging in small and heterogeneous populations with rare outcomes such as children with congenital heart disease. Finally, we review analytic considerations for the deployment of digital biomarkers in clinical settings, including the emerging field of clinical artificial intelligence (AI) operations, computational needs for deployment, efforts to increase the explainability of AI, algorithmic drift, and the needs for distributed surveillance and federated learning. We conclude this review by discussing a recent simulation study that shows that, despite these analytic challenges and complications, the use of digital biomarkers in managing clinical care might have substantial benefits regarding individual patient outcomes.