Predicting vancomycin exposure is essential for optimizing dosing regimens in sepsis patients. While population pharmacokinetic (PPK) models are commonly used, their performance is limited. Machine learning (ML) models offer advantages over PPK models, but it remains unclear which model-PPK, Bayesian, ML, or hybrid PPK-ML-is best for predicting vancomycin exposure across different clinical scenarios in sepsis patients. This study compares the performance of these models in predicting the 24 hour area under the blood concentration curve (AUC24) to support precision dosing in sepsis care. Data from sepsis patients treated with intravenous vancomycin were sourced from the MIMIC-IV database. The data set was split into training and testing sets, and four models-PPK, Bayesian, ML, and hybrid-were developed. In the testing set, AUC24 was predicted using all models, and performance was evaluated using mean absolute error, mean squared error, root mean squared error, mean absolute percentage error (MAPE), and R². A total of 4,059 patients were included. In the absence of vancomycin concentration data, the hybrid model outperformed both PPK and Bayesian models, with MAPE improvements of 58% and 17%, respectively. When vancomycin concentration data were available, the Bayesian model demonstrated the best performance (MAPE: 13.37% vs 68.17%, 34.17%, and 28.52% for PPK, Random Forest, and hybrid models). The hybrid model is recommended to predict AUC24 when concentration data were unavailable, while the Bayesian model should be used when concentrations were available, offering robust strategies for precise vancomycin dosing in sepsis patients.

This study evaluates and compares the performance of four models-PPK, Bayesian, ML, and hybrid PPK-ML-in predicting vancomycin exposure (AUC24) in sepsis patients using real-world data from the MIMIC-IV database. These results underscore the importance of selecting appropriate models based on the availability of concentration data, providing valuable guidance for precision dosing strategies in sepsis care. This work contributes to advancing personalized vancomycin therapy, optimizing dosing regimens, and improving clinical outcomes in sepsis patients.

Hydrocephalus, particularly congenital hydrocephalus in infants, remains underexplored in deep learning research. While deep learning has been widely applied to medical image analysis, few studies have specifically addressed the automated classification of hydrocephalus. This study proposes a convolutional neural network (CNN) model based on the VGG16 architecture to detect hydrocephalus in infant head CT images. The model integrates an automated method for ventricular volume extraction, applying windowing, histogram equalization, and thresholding techniques to segment the ventricles from surrounding brain structures. Morphological operations refine the segmentation and contours are extracted for visualization and volume measurement. The dataset consists of 105 head CT scans, each with 60 slices covering the ventricular volume, resulting in 6300 slices. Manual segmentation by three trained radiologists served as the reference standard. The automated method showed a high correlation with manual measurements, with R2 values ranging from 0.94 to 0.99. The mean absolute percentage error (MAPE) ranged 3.99 to 11.13%, while the root mean square error (RRMSE) from 4.56 to 13.74%. To improve model robustness, the dataset was preprocessed, normalized, and augmented with rotation, shifting, zooming, and flipping. The VGG16-based CNN used pre-trained convolutional layers with additional fully connected layers for classification, predicting hydrocephalus or normal labels. Performance evaluation using a multi-split strategy (15 independent splits) achieved a mean accuracy of 90.4% ± 1.2%. This study presents an automated approach for ventricular volume extraction and hydrocephalus detection, offering a promising tool for clinical and research applications with high accuracy and reduced observer bias.

Machine learning (ML), a major branch of artificial intelligence, has not only demonstrated the potential to significantly improve numerous sectors of healthcare but has also made significant contributions to the field of solid organ transplantation. ML provides revolutionary opportunities in areas such as donor-recipient matching, post-transplant monitoring, and patient care by automatically analyzing large amounts of data, identifying patterns, and forecasting outcomes.

To conduct a comprehensive bibliometric analysis of publications on the use of ML in transplantation to understand current research trends and their implications.

On July 18, a thorough search strategy was used with the Web of Science database. ML and transplantation-related keywords were utilized. With the aid of the VOS viewer application, the identified articles were subjected to bibliometric variable analysis in order to determine publication counts, citation counts, contributing countries, and institutions, among other factors.

Of the 529 articles that were first identified, 427 were deemed relevant for bibliometric analysis. A surge in publications was observed over the last four years, especially after 2018, signifying growing interest in this area. With 209 publications, the United States emerged as the top contributor. Notably, the "Journal of Heart and Lung Transplantation" and the "American Journal of Transplantation" emerged as the leading journals, publishing the highest number of relevant articles. Frequent keyword searches revealed that patient survival, mortality, outcomes, allocation, and risk assessment were significant themes of focus.

The growing body of pertinent publications highlights ML's growing presence in the field of solid organ transplantation. This bibliometric analysis highlights the growing importance of ML in transplant research and highlights its exciting potential to change medical practices and enhance patient outcomes. Encouraging collaboration between significant contributors can potentially fast-track advancements in this interdisciplinary domain.

Obstructive sleep apnea (OSA) is a prevalent sleep disorder with a high rate of undiagnosed patients, primarily due to the complexity of its diagnosis made by polysomnography (PSG). Considering the severe comorbidities associated with OSA, especially in the cardiovascular system, the development of early screening tools for this disease is imperative. Heart rate variability (HRV) is a simple and non-invasive approach used as a probe to evaluate cardiac autonomic modulation, with a variety of newly developed indices lacking studies with OSA patients.

We aimed to evaluate numerous HRV indices, derived from linear but mainly nonlinear indices, combined or not with oxygen saturation indices, for detecting the presence and severity of OSA using machine learning models.

ECG waveforms were collected from 291 PSG recordings to calculate 34 HRV indices. Minimum oxygen saturation value during sleep (SatMin), the percentage of total sleep time the patient spent with oxygen saturation below 90% (T90), and patient anthropometric data were also considered as inputs to the models. The Apnea-Hypopnea Index (AHI) was used to categorize into severity classes of OSA (normal, mild, moderate, severe) to train multiclass or binary (normal-to-mild and moderate-to-severe) classification models, using the Random Forest (RF) algorithm. Since the OSA severity groups were unbalanced, we used the Synthetic Minority Over-sampling Technique (SMOTE) to oversample the minority classes.

Multiclass models achieved a mean area under the ROC curve (AUROC) of 0.92 and 0.86 in classifying normal individuals and severe OSA patients, respectively, when using all attributes. When the groups were dichotomized into normal-to-mild OSA vs. moderate-to-severe OSA, an AUROC of 0.83 was obtained. As revealed by RF, the importance of features indicates that all feature modalities (HRV, SpO2, and anthropometric variables) contribute to the top 10 ranks.

The present study demonstrates the feasibility of using classification models to detect the presence and severity of OSA using these indices. Our findings have the potential to contribute to the development of rapid screening tools aimed at assisting individuals affected by this condition, to expedite diagnosis and initiate timely treatment.

Currently, there is a lack of effective early assessment tools for predicting the onset and development of cardiac arrest (CA). With the increasing attention of clinical researchers on machine learning (ML), some researchers have developed ML models for predicting the occurrence and prognosis of CA, with certain models appearing to outperform traditional scoring tools. However, these models still lack systematic evidence to substantiate their efficacy.

This systematic review and meta-analysis was conducted to evaluate the prediction value of ML in CA for occurrence, good neurological prognosis, mortality, and the return of spontaneous circulation (ROSC), thereby providing evidence-based support for the development and refinement of applicable clinical tools.

PubMed, Embase, the Cochrane Library, and Web of Science were systematically searched from their establishment until May 17, 2024. The risk of bias in all prediction models was assessed using the Prediction Model Risk of Bias Assessment Tool.

In total, 93 studies were selected, encompassing 5,729,721 in-hospital and out-of-hospital patients. The meta-analysis revealed that, for predicting CA, the pooled C-index, sensitivity, and specificity derived from the imbalanced validation dataset were 0.90 (95% CI 0.87-0.93), 0.83 (95% CI 0.79-0.87), and 0.93 (95% CI 0.88-0.96), respectively. On the basis of the balanced validation dataset, the pooled C-index, sensitivity, and specificity were 0.88 (95% CI 0.86-0.90), 0.72 (95% CI 0.49-0.95), and 0.79 (95% CI 0.68-0.91), respectively. For predicting the good cerebral performance category score 1 to 2, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.86 (95% CI 0.85-0.87), 0.72 (95% CI 0.61-0.81), and 0.79 (95% CI 0.66-0.88), respectively. For predicting CA mortality, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.85 (95% CI 0.82-0.87), 0.83 (95% CI 0.79-0.87), and 0.79 (95% CI 0.74-0.83), respectively. For predicting ROSC, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.77 (95% CI 0.74-0.80), 0.53 (95% CI 0.31-0.74), and 0.88 (95% CI 0.71-0.96), respectively. In predicting CA, the most significant modeling variables were respiratory rate, blood pressure, age, and temperature. In predicting a good cerebral performance category score 1 to 2, the most significant modeling variables in the in-hospital CA group were rhythm (shockable or nonshockable), age, medication use, and gender; the most significant modeling variables in the out-of-hospital CA group were age, rhythm (shockable or nonshockable), medication use, and ROSC.

ML represents a currently promising approach for predicting the occurrence and outcomes of CA. Therefore, in future research on CA, we may attempt to systematically update traditional scoring tools based on the superior performance of ML in specific outcomes, achieving artificial intelligence-driven enhancements.

PROSPERO International Prospective Register of Systematic Reviews CRD42024518949; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=518949.

One of the main goals of metagenomic studies is to describe the taxonomic diversity of microbial communities. A crucial step in metagenomic analysis is metagenomic binning, which involves the (supervised) classification or (unsupervised) clustering of metagenomic sequences. Various machine learning models have been applied to address this task. In this review, the contributions of artificial neural networks (ANN) in the context of metagenomic binning are detailed, addressing both supervised, unsupervised, and semi-supervised approaches. 34 ANN-based binning tools are systematically compared, detailing their architectures, input features, datasets, advantages, disadvantages, and other relevant aspects. The findings reveal that deep learning approaches, such as convolutional neural networks and autoencoders, achieve higher accuracy and scalability than traditional methods. Gaps in benchmarking practices are highlighted, and future directions are proposed, including standardized datasets and optimization of architectures, for third-generation sequencing. This review provides support to researchers in identifying trends and selecting suitable tools for the metagenomic binning problem.

The blood brain barrier (BBB) is an endothelial-derived structure which restricts the movement of certain molecules between the general somatic circulatory system to the central nervous system (CNS). While the BBB maintains homeostasis by regulating the molecular environment induced by cerebrovascular perfusion, it also presents significant challenges in developing therapeutics intended to act on CNS targets. Many drug development practices rely partly on extensive cell and animal models to predict, to an extent, whether prospective therapeutic molecules can cross the BBB. In interest to reduce costs and improve prediction accuracy, many propose using advanced computational modeling of BBB permeability profiles leveraging empirical data. Given the scale of growth in machine learning and deep learning, we review the most recent machine learning approaches in predicting BBB permeability.

Heart failure is a complex and prevalent condition with significant implications for patient management and survival prediction. Traditional predictive models often fall short in accuracy due to their reliance on pre-specified predictors and assumptions of variable independence. This review aims to assess the role of machine learning (ML) algorithms in predicting heart failure survival, comparing their performance with traditional statistical methods and identifying key predictive features. We conducted a review of studies utilizing ML algorithms for heart failure survival prediction. Data were sourced from PubMed/MEDLINE, Google Scholar, ScienceDirect, Embase, DOAJ, and the Cochrane Library, covering studies published until July 2024. A total of 10 studies were reviewed, encompassing 468,171 patients with heart failure. ML algorithms, particularly random forests and gradient boosting methods, demonstrated superior performance compared to traditional statistical models. These algorithms effectively identified key risk factors and stratified patients into risk categories with high accuracy. Notably, extreme learning machine (ELM) and CatBoost models showed exceptional predictive capabilities, as indicated by metrics such as Harrell's concordance index (C-index) and area under the curve (AUC). Key predictive features included ejection fraction (EF), serum creatinine (S Cr), and blood urea nitrogen (BUN). ML algorithms offer significant advantages in predicting heart failure survival by uncovering complex patterns and improving risk stratification. Their integration into clinical practice could lead to more personalized treatment strategies and enhanced patient outcomes. However, challenges such as data quality, model interpretability, and integration into clinical workflows need to be addressed.

Adeno-associated virus (AAV) has emerged as a leading vector for gene therapy due to its broad host range, low pathogenicity, and ability to facilitate long-term gene expression. However, AAV vectors face limitations, including immunogenicity and insufficient targeting specificity. To enhance the efficacy of gene therapy, researchers have been modifying the AAV vector using various methods. Traditional experimental approaches for optimizing AAV vector are often time-consuming, resource-intensive, and difficult to replicate. The advancement of artificial intelligence (AI), particularly machine learning, offers significant potential to accelerate capsid optimization while reducing development time and manufacturing costs. This review compares traditional and AI-based methods of AAV vector engineering and highlights recent research in AAV engineering using AI algorithms.

Auscultation is one of the key medical skills in physical examination. The main problem with auscultation is the lack of objectivity of the findings and great dependence on the experience of the examiner. Auscultation using machine learning and neural networks promises great potential for solving these problems in clinical practice.A selective search for studies in PubMed was carried out, which revealed the possibilities of machine learning in medical diagnostics.In all the studies identified, significant differences were shown between the respective test groups in favour of artificial intelligence (AI). In addition to the positive study results, the limitations of AI could also be analysed and critically scrutinised.Medical research in the field of artificial intelligence is still in its infancy. The prospects and limitations of AI must be further investigated and require close attention in the collaboration between clinicians, scientists and AI experts. Publicly funded projects such as DigitaLung (a digital auscultation system for the differential diagnosis of lung diseases using machine learning), which aims to improve lung auscultation using AI, will help to unlock the diagnostic benefits of AI for patient care and could improve care in the future.

We aimed to evaluate the image quality and diagnostic performance of chronic lung allograft dysfunction (CLAD) with lung ventilation single-photon emission computed tomography (SPECT) images acquired briefly using a convolutional neural network (CNN) in patients after lung transplantation and to explore the feasibility of short acquisition times. We retrospectively identified 93 consecutive lung-transplant recipients who underwent ventilation SPECT/computed tomography (CT). We employed a CNN to distinguish the images acquired in full time from those acquired in a short time. The image quality was evaluated using the structural similarity index (SSIM) loss and normalized mean square error (NMSE). The correlation between functional volume/morphological volume (F/M) ratios of full-time SPECT images and predicted SPECT images was evaluated. Differences in the F/M ratio were evaluated using Bland-Altman plots, and the diagnostic performance was compared using the area under the curve (AUC). The learning curve, obtained using MSE, converged within 100 epochs. The NMSE was significantly lower (P < 0.001) and the SSIM was significantly higher (P < 0.001) for the CNN-predicted SPECT images compared to the short-time SPECT images. The F/M ratio of full-time SPECT images and predicted SPECT images showed a significant correlation (r = 0.955, P < 0.0001). The Bland-Altman plot revealed a bias of -7.90% in the F/M ratio. The AUC values were 0.942 for full-time SPECT images, 0.934 for predicted SPECT images and 0.872 for short-time SPECT images. Our findings suggest that a deep-learning-based approach can significantly curtail the acquisition time of ventilation SPECT, while preserving the image quality and diagnostic accuracy for CLAD.

We aimed to characterize the brain abnormalities that are associated with the cognitive and physical performance of patients with relapsing-remitting multiple sclerosis (RRMS) using a deep learning algorithm.

Three-dimensional (3D) nnU-Net was employed to calculate a novel spatial abnormality map by T1-weighted images and 281 RRMS patients (Dataset-1, male/female = 101/180, median age [range] = 35.0 [17.0, 65.0] years) were categorized into subtypes. Comparison of clinical and MRI features between RRMS subtypes was conducted by Kruskal-Wallis test. Kaplan-Meier analysis was conducted to investigate disability progression in RRMS subtypes. Additional validation using two other RRMS datasets (Dataset-2, n = 33 and Dataset-3, n = 56) was conducted.

Five RRMS subtypes were identified: (1) a Frontal-I subtype showing preserved cognitive performance and mild physical disability, and low risk of disability worsening; (2) a Frontal-II subtype showing low cognitive scores and severe physical disability with significant brain volume loss, and a high propensity for disability worsening; (3) a temporal-cerebellar subtype demonstrating lowest cognitive scores and severest physical disability among all subtypes but remaining relatively stable during follow-up; (4) an occipital subtype demonstrating similar clinical and imaging characteristics as the Frontal-II subtype, except a large number of relapses at baseline and preserved cognitive performance; and (5) a subcortical subtype showing preserved cognitive performance and low physical disability but a similar prognosis as the occipital and Frontal-II subtypes. Additional validation confirmed the above findings.

Spatial abnormality maps can explain heterogeneity in cognitive and physical performance in RRMS and may contribute to stratified management.

Question Can a deep learning algorithm characterize the brain abnormalities associated with the cognitive and physical performance of patients with RRMS? Findings Five RRMS subtypes were identified by the algorithm that demonstrated variable cognitive and physical performance. Clinical relevance The spatial abnormality maps derived RRMS subtypes had distinct cognitive and physical performances, which have a potential for individually tailored management.

The atherogenic index of plasma (AIP) is considered an important marker of atherosclerosis and cardiovascular risk. However, its potential role in predicting length of stay (LOS), especially in patients with atherosclerotic cardiovascular disease (ASCVD), remains to be explored. We investigated the effect of AIP on hospital LOS in critically ill ASCVD patients and explored the risk factors affecting LOS in conjunction with machine learning.

Using data from the Medical Information Mart for Intensive Care (MIMIC)-IV. AIP was calculated as the logarithmic ratio of TG to HDL-C, and patients were stratified into four groups based on AIP values. We investigated the association between AIP and two key clinical outcomes: ICU LOS and total hospital LOS. Multivariate logistic regression models were used to evaluate these associations, while restricted cubic spline (RCS) regressions assessed potential nonlinear relationships. Additionally, machine learning (ML) techniques, including logistic regression (LR), decision tree (DT), random forest (RF), extreme gradient boosting (XGB), and light gradient boosting machine (LGB), were applied, with the Shapley additive explanation (SHAP) method used to determine feature importance.

The study enrolled a total of 2423 patients with critically ill ASCVD, predominantly male (54.91%), and revealed that higher AIP values were independently associated with longer ICU and hospital stays. Specifically, for each unit increase in AIP, the odds of prolonged ICU and hospital stays were significantly higher, with adjusted odds ratios (OR) of 1.42 (95% CI, 1.11-1.81; P = 0.006) and 1.73 (95% CI, 1.34-2.24; P < 0.001), respectively. The RCS regression demonstrated a linear relationship between increasing AIP and both ICU LOS and hospital LOS. ML models, specifically LGB (ROC:0.740) and LR (ROC:0.832) demonstrated superior predictive accuracy for these endpoints, identifying AIP as a vital component of hospitalization duration.

AIP is a significant predictor of ICU and hospital LOS in patients with critically ill ASCVD. AIP could serve as an early prognostic tool for guiding clinical decision-making and managing patient outcomes.

This study aimed to develop a clinical early warning prediction model to evaluate the prognosis and response to chemoimmunotherapy in patients with extensive-stage small cell lung cancer (ES-SCLC), thereby guiding clinical decision-making.

A retrospective analysis was conducted on the clinical data and radiomics parameters of 309 patients with ES-SCLC hospitalized at Baotou Cancer Hospital from February 2020 to September 2024. Patients were divided into reactive and non-reactive groups based on their response to chemoimmunotherapy.Machine learning algorithms (including random forests, decision trees, artificial neural networks, and generalized linear regression) were used to predict the combined treatment response. The model's predictive ability was evaluated using the receiver operating characteristic (ROC) curve and clinical decision curve analysis(DCA). The prognostic evaluation of patients receiving combination therapy was based on the COX regression model, with predictive performance assessed through nomogram visualization and calibration curves.

Out of 309 patients with ES-SCLC, 248 (80.26%) responded to combination therapy. Logistic regression and Least absolute shrinkage and selection operator (LASSO) regression analyses identified Energy, sum of squares(SOS), mean sum(MES), sum variance(SUV), sum entropy(SUE), difference variance(DIV), and pathomics score as independent risk factors for treatment response. The area under the ROC curve for predicting treatment response using machine learning were 0.764 (95% confidence interval [CI]: 0.707~0.821) and 0.901 (95% CI: 0.846~0.956) in the training and validation sets. The C-index of the radiomics and pathomics prognostic nomogram model based on the COX prognostic model was 0.766 and 0.812 in those sets, respectively.

We developed prediction model based on multi-omics demonstrated satisfactory performance in predicting chemoimmunotherapy response in patients with ES-SCLC. The random forest prediction model, in particular, provides accurate response and prognostic risk assessments, thereby assisting clinical decision-making.

Machine Learning (ML) has become an essential tool for analyzing biomedical data, facilitating the prediction of treatment outcomes and patient survival. However, the effectiveness of ML models heavily relies on both the choice of algorithms and the quality of the input data. In this study, we aimed to develop a novel predictive model to estimate individual survival for patients diagnosed with glioblastoma (GBM), focusing on key variables such as O6-Methylguanine-DNA Methyltransferase (MGMT) methylation status, age, and sex.

To identify the optimal approach, we utilized retrospective data from 218 patients treated at our brain tumor center. The performance of the ML models was evaluated within repeated tenfold regression. The pipeline comprised five regression estimators, including both linear and non-linear algorithms. Permutation feature importance highlighted the feature with the most significant impact on the model. Statistical significance was assessed using a permutation test procedure.

The best machine learning algorithm achieved a mean absolute error (MAE) of 12.65 (SD = ± 2.18) and an explained variance (EV) of 7% (SD = ± 1.8%) with p < 0.001. Linear algorithms led to more accurate predictions than non-linear estimators. Feature importance testing indicated that age and positive MGMT-methylation influenced the predictions the most.

In summary, here we provide a novel approach allowing to predict GBM patient's survival in months solely based on key parameters such as age, sex and MGMT-methylation status and underscores MGMT-methylation status as key prognostic factor for GBM patients survival.

The prognostic prediction of patients with hypercapnic respiratory failure holds significant clinical value. The objective of this study was to develop and validate a predictive model for predicting survival in patients with hypercapnic respiratory failure.

The study enrolled a total of 697 patients with hypercapnic respiratory failure, including 565 patients from the First People's Hospital of Yancheng in the modeling group and 132 patients from the People's Hospital of Jiangsu Province in the external validation group. The three selected models were random survival forest (RSF), DeepSurv, a deep learning-based survival prediction algorithm, and Cox Proportional Risk (CoxPH). The model's predictive performance was evaluated using the C-index and Brier score. Receiver operating characteristic curve (ROC), area under ROC curve (AUC), and decision curve analysis (DCA) were employed to assess the accuracy of predicting the prognosis for survival at 6, 12, 18, and 24 months.

The RSF model (c-index: 0.792) demonstrated superior predictive ability for the prognosis of patients with hypercapnic respiratory failure compared to both the traditional CoxPH model (c-index: 0.699) and DeepSurv model (c-index: 0.618), which was further validated on external datasets. The Brier Score of the RSF model demonstrated superior performance, consistently measuring below 0.25 at the 6-month, 12-month, 18-month, and 24-month intervals. The ROC curve confirmed the superior discrimination of the RSF model, while DCA demonstrated its optimal clinical net benefit in both the modeling group and the external validation group.

The RSF model offered distinct advantages over the CoxPH and DeepSurv models in terms of clinical evaluation and monitoring of patients with hypercapnic respiratory failure.

Diabetic peripheral neuropathy (DPN) is the most common complication of diabetes mellitus. Early identification of individuals at high risk of DPN is essential for successful early intervention. Traditional Chinese medicine (TCM) tongue diagnosis, one of the four diagnostic methods, lacks specific algorithms for TCM symptoms and tongue features. This study aims to develop machine learning (ML) models based on TCM to predict the risk of diabetic peripheral neuropathy (DPN) in patients with type 2 diabetes mellitus (T2DM).

A total of 4723 patients were included in the analysis (4430 with T2DM and 293 with DPN). TFDA-1 was used to obtain tongue images during a questionnaire survey. LASSO (least absolute shrinkage and selection operator) logistic regression model with fivefold cross-validation was used to select imaging features, which were then screened using best subset selection. The synthetic minority oversampling technique (SMOTE) algorithm was applied to address the class imbalance and eliminate possible bias. The area under the receiver operating characteristic curve (AUC) was used to evaluate the model's performance. Four ML algorithms, namely logistic regression (LR), random forest (RF), support vector classifier (SVC), and light gradient boosting machine (LGBM), were used to build predictive models for DPN. The importance of covariates in DPN was ranked using classifiers with better performance.

The RF model performed the best, with an accuracy of 0.767, precision of 0.718, recall of 0.874, F-1 score of 0.789, and AUC of 0.77. With a value of 0.879, the LGBM model appeared to be the best regarding recall Age, sweating, dark red tongue, insomnia, and smoking were the five most significant RF features. Age, yellow coating, loose teeth, smoking, and insomnia were the five most significant features of the LGBM model.

This cross-sectional study demonstrates that the RF and LGBM models can screen for high-risk DPN in T2DM patients using TCM symptoms and tongue features. The identified key TCM-related features, such as age, tongue coating, and other symptoms, may be advantageous in developing preventative measures for T2DM patients.

Dietary intervention plays a vital role in improving the prognosis of people with diabetes mellitus (DM). Currently, there is a lack of systematic analysis of the relation between dietary nutrients and long-term mortality risk in people with DM. The study aims to establish models predicting long-term mortality and explore dietary nutrients associated with reduced long-term events to guide daily dietary decisions in people with DM. The retrospective cohort study collected 5060 participants with DM from the National Health and Nutrition Examination Survey (NHANES) 1999-2018. The least absolute shrinkage and selection operator (LASSO) regression and random forest (RF) algorithm were applied to identify key mortality-related dietary factors, which were subsequently incorporated into risk prediction nomogram models. The receiver operating characteristic (ROC) curve, calibration plot and decision curve analysis (DCA) were utilized to evaluate the performance of the models. The association of key dietary nutrients with all-cause and cardiovascular mortality were visualized by restricted cubic spline (RCS) models both in the whole and subgroups by sex, age, drinking and smoking status. The overall median age of the cohort was 62.0 years (interquartile range (IQR) 52.0-70.0), 2564 (50.67%) being male. During a median follow-up period of 56.0 months, 997 (19.70%) all-cause deaths were recorded, with 219 (21.97%) of which being ascribed to cardiovascular disease. The nomogram models based on key dietary nutrients identified by LASSO and RF demonstrated a significant predicative value for all-cause and cardiovascular mortality. Dietary fiber and magnesium were the common predictive nutrients in the two nomogram models. The RCS curve revealed that dietary fiber and magnesium were negatively associated with long-term mortality in the whole and subgroups of people with DM after adjustment of potential confounders. The diet of people with DM is closely associated with mortality. The nomogram models based on dietary nutrients can predict long-term mortality of people with DM, and the higher intake of dietary fiber and magnesium was associated with reduced risks of both long-term all-cause and cardiovascular mortality.

The use of augmented or artificial intelligence (AI) in healthcare promises groundbreaking advancements, from increasing diagnostic accuracy and minimizing clinical errors to personalized treatment plans and automated clinical decision-making. Its use may allow us to transition from phenomenological categories of psychiatric illness to one driven by underlying etiology and realize the Research Domain Criteria (RDoC) model proposed by the (U.S.) National Institutes of Mental Health (NIMH), which today remains difficult to apply clinically and is accessible primarily to researchers. AI may facilitate the transition to a more syncretic framework of understanding psychiatric illness that accounts for disruptions, all the way from the cellular level to the level of social systems. Yet, despite immense possibilities, there are also associated risks. In this article, we explore the challenges and opportunities associated with the use of AI in psychiatry, focusing on the potential ethical and health equity considerations in vulnerable populations, especially in child and adolescent psychiatry.

In recent years, a sharp increase in artificial intelligence (AI)-based software as medical devices has been seen in the United States and the European Union. Despite the huge potential of these devices in alleviating suffering through rapid identification and early intervention, their adoption in clinical practice has remained relatively slow due to ethical questions surrounding their usage. Even though there is no universal framework for the approval of these devices, the guiding principles behind individual regulatory bodies almost stay the same, with some more focused on the technical aspect while others involving the ethical aspects as well. The International Medical Device Regulators Forum devised a SaMD Working Group to outline the essential controls guiding the approval of these devices, but there is a lack of a structured approach for the regulatory approval process. This article outlines the principles of medical ethics, such as autonomy, beneficence, and fair distribution of healthcare sources, and how they relate to the use of AI-based devices. The core regulatory guidelines are then viewed in light of these ethical principles. We recommend that a comprehensive regulatory framework with integration of principles of medical ethics be made public. Though no universally accepted framework is available, regulating quality management, risk assessment, and data privacy would help build trust to promote the adoption of AI in healthcare.

With the rising diagnostic rate of gallbladder polypoid lesions (GPLs), differentiating benign cholesterol polyps from gallbladder adenomas with a higher preoperative malignancy risk is crucial. This study aimed to establish a preoperative prediction model capable of accurately distinguishing between gallbladder adenomas and cholesterol polyps using machine learning algorithms.

We retrospectively analysed the patients' clinical baseline data, serological indicators, and ultrasound imaging data. Using 12 machine learning algorithms, 110 combination predictive models were constructed. The models were evaluated using internal and external cohort validation, receiver operating characteristic curves, area under the curve (AUC) values, calibration curves, and clinical decision curves to determine the best predictive model.

Among the 110 combination predictive models, the Support Vector Machine + Random Forest (SVM + RF) model demonstrated the highest AUC values of 0.972 and 0.922 in the training and internal validation sets, respectively, indicating an optimal predictive performance. The model-selected features included gallbladder wall thickness, polyp size, polyp echo, and pedicle. Evaluation through external cohort validation, calibration curves, and clinical decision curves further confirmed its excellent predictive ability for distinguishing gallbladder adenomas from cholesterol polyps. Additionally, this study identified age, adenosine deaminase level, and metabolic syndrome as potential predictive factors for gallbladder adenomas.

This study employed the machine learning combination algorithms and preoperative ultrasound imaging data to construct an SVM + RF predictive model, enabling effective preoperative differentiation of gallbladder adenomas and cholesterol polyps. These findings will assist clinicians in accurately assessing the risk of GPLs and providing personalised treatment strategies.

This study aimed to evaluate the predictive performance of inflammatory and nutritional indices for adverse cardiovascular events (ACE) in patients with acute myocardial infarction (AMI) after percutaneous coronary intervention (PCI) using a machine learning (ML) algorithm.

AMI patients who underwent PCI were recruited and randomly divided into non/ACE groups. Inflammatory and nutritional indices were graded according to the laboratory examination reports. Logistic Regression was used to screen for factors that were significant for ML model establishment. The performances of the algorithms were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc. RESULTS: Age, LVEF%, Killip Grade, heart rate, creatinine, albumin, neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and prognostic nutritional index (PNI) were significantly correlated with ACE by Logistic regression analysis (P < 0.05). These nine factors were employed to establish stepwise regression (SR), random forest (RF), naïve Bayes (NB), decision trees (DT), and artificial neutron network (ANN), whose performances were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc. The accuracy of the decision tree was greater than that of other trees. The area under the curves was the highest in the ANN model compared with the other models.

ANN predictive performance had an advantage over other ML algorithms based on age, LVEF%, Killip Grade, heart rate, creatinine, albumin, NLR, PLR, and PNI.

Although more risk prediction models are available for feeding intolerance in enteral-nourishment patients, it is still unclear how well these models will work in clinical settings. Future research faces challenges in validating model accuracy across populations, enhancing interpretability for clinical use, and overcoming dataset limitations.

To thoroughly examine studies that have been published on feeding intolerance risk prediction models for enteral nutrition patients.

Conducted a systematic review and meta-analysis of observational studies.

A comprehensive search of the literature was conducted using a range of databases, including China National Knowledge Infrastructure (CNKI), Wanfang Database, China Science and Technology Journal Database (VIP), SinoMed, PubMed, Web of Science, The Cochrane Library, Cumulative Index to Nursing and Allied Health Literature (CINAHL) and Embase. The search scope was confined to articles within the database from its inception until August 12th, 2024. The data from the selected studies should be extracted, including study design, subjects, duration of follow-up, data sources, outcome measures, sample size, handling of missing data, continuous variable handling methods, variable selection, final predictors, model development and performance, and form of model presentation. The applicability and bias risk were evaluated using the Prediction Model Risk of Bias Assessment Tool (PROBAST) checklist.

A total of 1,472 studies were retrieved. Following the selection criteria, 18 prediction models sourced from 14 studies were incorporated into this review. In the field of model construction, only one study employed the use of multiple machine-learning techniques for the development of a model. In contrast, the remaining studies used logistic regression to construct FI risk prediction models. The incidence of FI in enteral nutrition was 32.4-63.1%. The top five predictors included in the model were APACHE II, age, albumin levels, intra-abdominal pressure, and mechanical ventilation. The reported AUC, or area under the curve, exhibited a range of values between 0.70 and 0.921. All studies were identified as having a high risk of bias, primarily due to the use of inappropriate data sources and inadequate reporting within the analysis domain.

Although the included studies reported a certain degree of discriminatory power in their predictive models to identify feeding intolerance in patients undergoing enteral nutrition, the PROBAST assessment tool deemed all the included studies to carry a significant risk of bias. Future research should emphasize the development of innovative predictive models. These endeavors should incorporate more extensive and diverse sample sizes, adhere to stringent methodological designs, and undergo rigorous multicenter external validation to ensure robustness and generalizability.

Identifier CRD42024585099, https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=585099.

Using machine learning to construct a prediction model for the risk of diabetes kidney disease (DKD) in the American diabetes population and evaluate its effect.

First, a dataset of five cycles from 2009 to 2018 was obtained from the National Health and Nutrition Examination Survey (NHANES) database, weighted and then standardized (with the study population in the United States), and the data were processed and randomly grouped using R software. Next, variable selection for DKD patients was conducted using Lasso regression, two-way stepwise iterative regression, and random forest methods. A nomogram model was constructed for the risk prediction of DKD. Finally, the predictive performance, predictive value, calibration, and clinical effectiveness of the model were evaluated through the receipt of ROC curves, Brier score values, calibration curves (CC), and decision curves (DCA). In addition, we will visualize it.

A total of 4371 participants were selected and included in this study. Patients were randomly divided into a training set (n = 3066 people) and a validation set (n = 1305 people) in a 7:3 ratio. Using machine learning algorithms and drawing Venn diagrams, five variables significantly correlated with DKD risk were identified, namely Age, Hba1c, ALB, Scr, and TP. The area under the ROC curve (AUC) of the training set evaluation index for this model is 0.735, the net benefit rate of DCA is 2%-90%, and the Brier score is 0.172. The area under the ROC curve of the validation set (AUC) is 0.717, and the DCA curve shows a good net benefit rate. The Brier score is 0.177, and the calibration curve results of the validation set and training set are almost consistent.

The DKD risk nomogram model constructed in this study has good predictive performance, which helps to evaluate the risk of DKD as early as possible in clinical practice and formulate relevant intervention and treatment measures. The visual result can be used by doctors or individuals to estimate the probability of DKD risk, as a reference to help make better treatment decisions.

Gastric cancer (GC) has a poor prognosis, considerable cellular heterogeneity, and ranks fifth among malignant tumours. Understanding the tumour microenvironment (TME) and intra-tumor heterogeneity (ITH) may lead to the development of novel GC treatments.

The single-cell RNA sequencing (scRNA-seq) dataset was obtained from the Gene Expression Omnibus (GEO) database, where diverse immune cells were isolated and re-annotated based on cell markers established in the original study to ascertain their individual characteristics. We conducted a weighted gene co-expression network analysis (WGCNA) to identify genes with a significant correlation to GC. Utilising bulk RNA sequencing data, we employed machine learning integration methods to train specific biomarkers for the development of novel diagnostic combinations. A two-sample Mendelian randomisation study was performed to investigate the causal effect of biomarkers on gastric cancer (GC). Ultimately, we utilised the DSigDB database to acquire associations between signature genes and pharmaceuticals.

The 18 genes that made up the signature were as follows: ZFAND2A, PBX4, RAMP2, NNMT, RNASE1, CD93, CDH5, NFKBIE, VWF, DAB2, FAAH2, VAT1, MRAS, TSPAN4, EPAS1, AFAP1L1, DNM3. Patients were categorised into high-risk and low-risk groups according to their risk scores. Individuals in the high-risk cohort exhibited a dismal outlook. The Mendelian randomisation study demonstrated that individuals with a genetic predisposition for elevated NFKBIE levels exhibited a heightened likelihood of acquiring GC. Molecular docking indicates that gemcitabine and chloropyramine may serve as effective therapeutics against NFKBIE.

We developed and validated a signature utilising scRNA-seq and bulk sequencing data from gastric cancer patients. NFKBIE may function as a novel biomarker and therapeutic target for GC.

The identification of Aucklandiae Radix (AR), Vladimiriae Radix (VR), and Inulae Radix (IR) based on traits and microscopic features is susceptible to the state of samples and the subjective awareness of personnel, and the identification based on a few or single chemical compositions is a cumbersome and time-consuming procedure and fails to rationally and effectively utilize the information of unknown components and is not specificity enough.

This study aimed to improve the identification efficiency, strengthen supervision, and realize digital identification of three Chinese medicines. Ultra-high-performance liquid chromatography with quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) combined with multivariate algorithms was used to explore the digital identification of AR, VR, and IR.

UHPLC-QTOF-MS was used to analyze AR, VR, and IR. The MS data combined with multivariate algorithms such as partial least squares discrimination analysis (PLS-DA) and artificial neural networks (ANNs) was used to filter important variables and data modeling. Finally, the optimal model was selected for the digital identification of three herbs.

The results showed that three herbs can be distinguished on the whole level, and through feature screening, 591 characteristic variables combined with multivariate algorithms to construct data models. The ANN model was the best with accuracy = 0.983, precision = 0.984, and external verification showed the reliability and practicability of ANN model.

ANN model combined with MS data is of great significance for tdigital identification of AR, VR, and IR. It is an important reference for developing the digital identification of traditional Chinese medicines at the individual level based on UHPLC-QTOF-MS and multivariate algorithms.

To summarise the currently developed risk prediction models for medication adherence in patients with chronic diseases and evaluate their performance and applicability.

Ensuring medication adherence is crucial in effectively managing chronic diseases. Although numerous studies have endeavoured to construct risk prediction models for predicting medication adherence in patients with chronic illnesses, the reliability and practicality of these models remain uncertain.

Systematic review.

We conducted searches on PubMed, Web of Science, Cochrane, CINAHL, Embase and Medline from inception until 16 July 2023. Two authors independently screened risk prediction models for medication adherence that met the predefined inclusion criteria. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was employed to evaluate both the risk of bias and clinical applicability of the included studies. This systematic review adhered to the 2020 PRISMA checklist.

The study included a total of 11 risk prediction models from 11 studies. Medication regimen and age were the most common predictors. The use of PROBAST revealed that some essential methodological details were not thoroughly reported in these models. Due to limitations in methodology, all models were rated as having a high-risk for bias.

According to PROBAST, the current models for predicting medication adherence in patients with chronic diseases exhibit a high risk of bias. Future research should prioritise enhancing the methodological quality of model development and conducting external validations on existing models.

Based on the review findings, recommendations have been provided to refine the construction methodology of prediction models with an aim of identifying high-risk individuals and key factors associated with low medication adherence in chronic diseases.

This systematic review was conducted without patient or public participation.

While public databases like Transfermarkt provide valuable data for assessing the impact of anterior cruciate ligament (ACL) injuries in professional footballers, they require robust verification methods due to accuracy concerns. We hypothesised that an artificial intelligence (AI)-powered framework could cross-check ACL tear-related information from large publicly available data sets with high specificity.

The AI-powered framework uses Google Programmable Search Engine to search a curated, multilingual list of websites and OpenAI's GPT to translate search queries, appraise search results and analyse injury-related information in search result items (SRIs). Specificity was the chosen performance metric-the AI-powered framework's ability to accurately identify texts that do not mention an athlete suffering an ACL tear-with SRI as the evaluation unit. A database of ACL tears in male professional footballers from first- and second-tier leagues worldwide (1999-2024) was collected from Transfermarkt.com, and players were randomly selected for appraisal until enough SRIs were obtained to validate the framework's specificity. Player age at injury and time until return-to-play (RTP) were recorded and compared with Union of European Football Associations (UEFA) Elite Club Injury Study data.

Verification of 231 athletes yielded 1546 SRIs. Human analysis of the SRIs showed that 335 mentioned an ACL tear, corresponding to 83 athletes with ACL tears. Specificity and sensitivity of GPT in identifying mentions of ACL tears in a player were 99.3% and 88.4%, respectively. Mean age at rupture was 26.6 years (standard deviation: 4.6, 95% confidence interval [CI]: 25.6-27.6). Median RTP time was 225 days (interquartile range: 96, 95% CI: 209-251), which is comparable to reports using data from the UEFA Elite Club Injury Study.

This study shows that an AI-powered framework can achieve high specificity in cross-checking ACL tear reports in male professional football from public databases, markedly reducing manual workload and enhancing the reliability of media-based sports medicine research.

Level III.

Lumbar spinal stenosis (LSS) is a major cause of pain and disability in older individuals worldwide. Although increasing studies of traditional machine learning (TML) and deep learning (DL) were conducted in the field of diagnosing LSS and gained prominent results, the performance of these models has not been analyzed systematically.

This systematic review and meta-analysis aimed to pool the results and evaluate the heterogeneity of the current studies in using TML or DL models to diagnose LSS, thereby providing more comprehensive information for further clinical application.

This review was performed under the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines using articles extracted from PubMed, Embase databases, and Cochrane Library databases. Studies that evaluated DL or TML algorithms assessment value on diagnosing LSS were included, while those with duplicated or unavailable data were excluded. Quality Assessment of Diagnostic Accuracy Studies 2 was used to estimate the risk of bias in each study. The MIDAS module and the METAPROP module of Stata (StataCorp) were used for data synthesis and statistical analyses.

A total of 12 studies with 15,044 patients reported the assessment value of TML or DL models for diagnosing LSS. The risk of bias assessment yielded 4 studies with high risk of bias, 3 with unclear risk of bias, and 5 with completely low risk of bias. The pooled sensitivity and specificity were 0.84 (95% CI: 0.82-0.86; I2=99.06%) and 0.87 (95% CI 0.84-0.90; I2=98.7%), respectively. The diagnostic odds ratio was 36 (95% CI 26-49), the positive likelihood ratio (LR+) was 6.6 (95% CI 5.1-8.4), and the negative likelihood ratio (LR-) was 0.18 (95% CI 0.16-0.21). The summary receiver operating characteristic curves, the area under the curve of TML or DL models for diagnosing LSS of 0.92 (95% CI 0.89-0.94), indicating a high diagnostic value.

This systematic review and meta-analysis emphasize that despite the generally satisfactory diagnostic performance of artificial intelligence systems in the experimental stage for the diagnosis of LSS, none of them is reliable and practical enough to apply in real clinical practice. Further efforts, including optimization of model balance, widely accepted objective reference standards, multimodal strategy, large dataset for training and testing, external validation, and sufficient and scientific report, should be made to bridge the distance between current TML or DL models and real-life clinical applications in future studies.

PROSPERO CRD42024566535; https://tinyurl.com/msx59x8k.

Colorectal cancer (CRC) is characterized by high heterogeneity, aggressiveness, and high morbidity and mortality rates. With machine learning (ML) algorithms, patient, tumor, and treatment features can be used to develop and validate models for predicting survival. In addition, important variables can be screened and different applications can be provided that could serve as vital references when making clinical decisions and potentially improving patient outcomes in clinical settings.

To construct prognostic prediction models and screen important variables for patients with stage I to III CRC.

More than 1000 postoperative CRC patients were grouped according to survival time (with cutoff values of 3 years and 5 years) and assigned to training and testing cohorts (7:3). For each 3-category survival time, predictions were made by 4 ML algorithms (all-variable and important variable-only datasets), each of which was validated via 5-fold cross-validation and bootstrap validation. Important variables were screened with multivariable regression methods. Model performance was evaluated and compared before and after variable screening with the area under the curve (AUC). SHapley Additive exPlanations (SHAP) further demonstrated the impact of important variables on model decision-making. Nomograms were constructed for practical model application.

Our ML models performed well; the model performance before and after important parameter identification was consistent, and variable screening was effective. The highest pre- and postscreening model AUCs 95% confidence intervals in the testing set were 0.87 (0.81-0.92) and 0.89 (0.84-0.93) for overall survival, 0.75 (0.69-0.82) and 0.73 (0.64-0.81) for disease-free survival, 0.95 (0.88-1.00) and 0.88 (0.75-0.97) for recurrence-free survival, and 0.76 (0.47-0.95) and 0.80 (0.53-0.94) for distant metastasis-free survival. Repeated cross-validation and bootstrap validation were performed in both the training and testing datasets. The SHAP values of the important variables were consistent with the clinicopathological characteristics of patients with tumors. The nomograms were created.

We constructed a comprehensive, high-accuracy, important variable-based ML architecture for predicting the 3-category survival times. This architecture could serve as a vital reference for managing CRC patients.

Ovarian cancer (OC) remains one of the most lethal gynecological malignancies globally. Despite the implementation of various medical imaging approaches for OC screening, achieving accurate differential diagnosis of ovarian tumors continues to pose significant challenges due to variability in image performance, resulting in a lack of objectivity that relies heavily on the expertise of medical professionals. This challenge can be addressed through the emergence and advancement of radiomics, which enables high-throughput extraction of valuable information from conventional medical images. Furthermore, radiomics can integrate with genomics, a novel approach termed radiogenomics, which allows for a more comprehensive, precise, and personalized assessment of tumor biological features. In this review, we present an extensive overview of the application of radiomics and radiogenomics in diagnosing and predicting ovarian tumors. The findings indicate that artificial intelligence methods based on imaging can accurately differentiate between benign and malignant ovarian tumors, as well as classify their subtypes. Moreover, these methods are effective in forecasting survival rates, treatment outcomes, metastasis risk, and recurrence for patients with OC. It is anticipated that these advancements will function as decision-support tools for managing OC while contributing to the advancement of precision medicine.

Distant metastasis in bladder cancer is linked to poor prognosis and significant mortality. Machine learning (ML), a key area of artificial intelligence, has shown promise in the diagnosis, staging, and treatment of bladder cancer. This study aimed to employ various ML techniques to predict distant metastasis in patients with bladder cancer.

Patients diagnosed with bladder cancer in the Surveillance, Epidemiology, and End Results (SEER) database from 2000 to 2021 were included in this study. After a rigorous screening process, a total of 4,108 patients were selected for further analysis, divided in a 7:3 ratio into a training cohort and an internal validation cohort. In addition, 118 patients treated at the Second Affiliated Hospital of Nanchang University were included as an external validation cohort. Features were filtered using the least absolute shrinkage and selection operator (LASSO) regression algorithm. Based on the significant features identified, three ML algorithms were utilized to develop prediction models: logistic regression, support vector machine (SVM), and linear discriminant analysis (LDA). The predictive performance of the three models was evaluated by obtaining the area under the receiver operating characteristic (ROC) curve (AUC), the precision, the accuracy, and the F1 score.

According to the statistical results, the final probability of distant metastasis in the population was 12.0% (n = 495). LASSO regression analysis revealed that age, chemotherapy, tumor size, the examination of non-regional lymph nodes, and regional lymph node evaluation were significantly associated with distant metastasis of bladder cancer. In the internal validation cohort, the prediction accuracy rates for logistic regression, SVM, and LDA were 0.874, 0.877, and 0.845, respectively. The precision rates were 0.805, 0.769, and 0.827, respectively, and the F1 scores were 0.821, 0.819, and 0.835, respectively. The ROC curve demonstrated that the AUC for all models was greater than 0.7. In the external validation cohort, the prediction accuracy rates for logistic regression, SVM, and LDA were 0.856, 0.848, and 0.797, respectively, with the ROC curve indicating that the AUC also exceeded 0.7. The precision rates were 0.877, 0.718, and 0.736, respectively, and the F1 scores were 0.797, 0.778, and 0.762, respectively. Among the algorithms used, logistic regression demonstrated better predictive efficiency than the other two methods. The top three variables with the highest importance scores in the logistic regression were non-regional lymph nodes, age, and chemotherapy.

The prediction model developed using three ML algorithms demonstrated strong accuracy and discriminative capability in predicting distant metastasis in patients with bladder cancer. This might help clinicians in understanding patient prognosis and in formulating personalized treatment strategies, ultimately improving the overall prognosis of patients with bladder cancer.

Obstructive sleep apnoea (OSA) is a prevalent sleep disorder characterized by pharyngeal airway collapse during sleep, leading to intermittent hypoxia, intrathoracic pressure swings, and sleep fragmentation. OSA is associated with various comorbidities and risk factors, contributing to its substantial economic and social burden. Machine learning (ML) techniques offer promise in predicting OSA severity and understanding its complex pathogenesis. This study aims to compare the accuracy of different ML techniques in predicting OSA severity and identify key associated factors contributing to OSA.

Adult patients suspected of OSA underwent clinical assessments and polysomnography. Demographic, anthropometric and clinical data were collected. Five supervised ML models (logistic regression, decision tree, random forest, extreme gradient boosting, support vector machine) were employed, optimized through grid search and cross-validation.

ML models exhibited varied performance across OSA severity levels. SVM demonstrated the highest accuracy for mild OSA, XGBoost for moderate OSA, and random forest for severe OSA. Logistic regression showed the highest AUC for moderate and severe OSA. Anthropometric measures, gender, and hypertension were significant predictors of OSA severity.

ML models offer valuable insights into predicting OSA severity and identifying associated factors. Our findings support the relevant potential clinical utility of ML in OSA management, although further validation and refinement are warranted.

The varying growth rates within a group of pigs present a significant challenge for the current all-in-all-out systems in the pig industry. This study evaluated the applicability of statistical methods for classifying pigs at risk of growth retardation at different production stages using a robust dataset collected under commercial conditions. Data from 26,749 crossbred pigs (Yorkshire × Landrace) with Duroc at weaning (17 to 27 d), 15,409 pigs at the end of the nursery period (60 to 78 d), and 4996 pigs at slaughter (151 to 161 d) were analyzed under three different cut points (lowest 10%, 20%, and 30% weights) to characterize light animals. Records were randomly split into training and testing sets in a 2:1 ratio, and each training dataset was analyzed using an ordinary least squares approach and three machine learning algorithms (decision tree, random forest, and generalized boosted regression). The classification performance of each analytical approach was evaluated by the area under the curve (AUC). In all production stages and cut points, the random forest and generalized boosted regression models demonstrated superior classification performance, with AUC estimates ranging from 0.772 to 0.861. The parametric linear model also showed acceptable classification performance, with slightly lower AUC estimates ranging from 0.752 to 0.818. In contrast, the single decision tree was categorized as worthless, with AUC estimates between 0.608 and 0.726. Key prediction factors varied across production stages, with birthweight-related factors being most significant at weaning, and weight at previous stages becoming more crucial later in the production cycle. These findings suggest the potential of machine learning algorithms to improve decision-making and efficiency in pig production systems by accurately identifying pigs at risk of growth retardation.

Retinal vasculitis (RV) is characterised by retinal vascular leakage, occlusion or both on fluorescein angiography (FA). There is no standard scheme available to segment RV features. We aimed to develop a deep learning model to segment both vascular leakage and occlusion in RV.

Four hundred and sixty-three FA images from 82 patients with retinal vasculitis were used to develop a deep learning model, in 60:20:20 ratio for training:validation:testing. Parameters, including deep learning architectures (DeeplabV3+, UNet++ and UNet), were altered to find the best binary segmentation model separately for retinal vascular leakage and occlusion, using a Dice score to determine the reliability of each model.

Our best model for vascular leakage had a Dice score of 0.6279 (95% confidence interval (CI) 0.5584-0.6974). For occlusion, the best model achieved a Dice score of 0.6992 (95% CI 0.6109-0.7874).

Our RV segmentation models could perform reliable segmentation for retinal vascular leakage and occlusion in FAs of RV patients.

To explore the differences and associations of hypoxic parameters among distinct types of respiratory events in patients with obstructive sleep apnea (OSA) and to construct prediction models for the types of respiratory events based on hypoxic parameters.

A retrospective analysis was conducted on a cohort of 67 patients with polysomnography (PSG). All overnight recorded respiratory events with pulse oxygen saturation (SpO2) desaturation were categorized into four categories: hypopnea (Hyp, 3409 events), obstructive apnea (OA, 5561 events), central apnea (CA, 1110 events) and mixed apnea (MA, 1372 events). All event recordings were exported separately from the PSG software as comma-separated variable (.csv) files, which were imported into custom-built MATLAB software for analysis. Based on 13 hypoxic parameters, artificial neural network (ANN) and binary logistic regression (BLR) were separately used for construction of Hyp, OA, CA and MA models. Receiver operating characteristic (ROC) curves were employed to compare the various predictive indicators of the two models for different respiratory event types, respectively.

Both ANN and BLR models suggested that 13 hypoxic parameters significantly influenced the classification of respiratory event types; The area under the ROC curves of the ANN models surpassed those of traditional BLR models respiratory event types.

The ANN models constructed based on the 13 hypoxic parameters exhibited superior predictive capabilities for distinct types of respiratory events, providing a feasible new tool for automatic identification of respiratory event types using sleep SpO2.

Recurrence for high-grade gliomas is inevitable despite maximal safe resection and adjuvant chemoradiation, and current imaging techniques fall short in predicting future progression. However, we introduce a novel whole-brain magnetic resonance spectroscopy (WB-MRS) protocol that delves into the intricacies of tumor microenvironments, offering a comprehensive understanding of glioma progression to inform expectant surgical and adjuvant intervention.

We investigated five locoregional tumor metabolites in a post-treatment population and applied machine learning (ML) techniques to analyze key relationships within seven regions of interest: contralateral normal-appearing white matter (NAWM), fluid-attenuated inversion recovery (FLAIR), contrast-enhancing tumor at time of WB-MRS (Tumor), areas of future recurrence (AFR), whole-brain healthy (WBH), non-progressive FLAIR (NPF), and progressive FLAIR (PF). Five supervised ML classification models and a neural network were developed, optimized, trained, tested, and validated. Lastly, a web application was developed to host our novel calculator, the Miami Glioma Prediction Map (MGPM), for open-source interaction.

Sixteen patients with histopathological confirmation of high-grade glioma prior to WB-MRS were included in this study, totaling 118,922 whole-brain voxels. ML models successfully differentiated normal-appearing white matter from tumor and future progression. Notably, the highest performing ML model predicted glioma progression within fluid-attenuated inversion recovery (FLAIR) signal in the post-treatment setting (mean AUC = 0.86), with Cho/Cr as the most important feature.

This study marks a significant milestone as the first of its kind to unveil radiographic occult glioma progression in post-treatment gliomas within 8 months of discovery. These findings underscore the utility of ML-based WB-MRS growth predictions, presenting a promising avenue for the guidance of early treatment decision-making. This research represents a crucial advancement in predicting the timing and location of glioblastoma recurrence, which can inform treatment decisions to improve patient outcomes.

Machine learning (ML) has been widely applied to predict the outcomes of numerous diseases. The current study aimed to develop a prognostic prediction model using machine learning algorithms and identify risk factors associated with residual back pain in patients with osteoporotic vertebrae compression fracture (OVCF) following percutaneous vertebroplasty (PVP).

A total of 863 OVCF patients who underwent PVP surgery were enrolled and analyzed. One month following surgery, a Visual Analog Scale (VAS) score of ≥ 4 was deemed to signify residual low back pain following the operation and patients were grouped into a residual pain group and pain-free group. The optimal feature set for both machine learning and statistical models was adjusted based on a 2000-resample bootstrap-based internal validation via an exhaustive search. The area under the curve (AUC), classification accuracy, sensitivity, and specificity of each model were then calculated to evaluate the predictive performance of each model.

In our current study, two main findings were observed: (1) Compared with statistical models, ML models exhibited superior predictive performance, with SVM demonstrating the highest prediction accuracy; (2) several variables were identified as the most predictive factors by both the machine learning and statistical models, including bone cement volume, number of fractured vertebrae, facet joint violation, paraspinal muscle degeneration, and intravertebral vacuum cleft.

Overall, the study demonstrated that machine learning classifiers such as SVM can effectively predict residual back pain for patients with OVCF following PVP while identifying associated predictors in a multivariate manner.