It is not rare to experience a biochemical recurrence (BCR) following radical prostatectomy (RP) for prostate cancer (PCa). It has been reported that early detection and management of BCR following surgery could improve survival in PCa. This study aimed to develop a nomogram integrating deep learning-based radiomic features and clinical parameters to predict 1-year BCR after RP and to examine the associations between radiomic scores and the tumor microenvironment (TME). In this retrospective multicenter study, two independent cohorts of patients (n = 349) who underwent RP after multiparametric magnetic resonance imaging (mpMRI) between January 2015 and January 2022 were included in the analysis. Single-cell RNA sequencing data from four prospectively enrolled participants were used to investigate the radiomic score-related TME. The 3D U-Net was trained and optimized for prostate cancer segmentation using diffusion-weighted imaging, and radiomic features of the target lesion were extracted. Predictive nomograms were developed via multivariate Cox proportional hazard regression analysis. The nomograms were assessed for discrimination, calibration, and clinical usefulness. In the development cohort, the clinical-radiomic nomogram had an AUC of 0.892 (95% confidence interval: 0.783--0.939), which was considerably greater than those of the radiomic signature and clinical model. The Hosmer-Lemeshow test demonstrated that the clinical-radiomic model performed well in both the development (P = 0.461) and validation (P = 0.722) cohorts. Decision curve analysis revealed that the clinical-radiomic nomogram displayed better clinical predictive usefulness than the clinical or radiomic signature alone in both cohorts. Radiomic scores were associated with a significant difference in the TME pattern. Our study demonstrated the feasibility of a DWI-based clinical-radiomic nomogram combined with deep learning for the prediction of 1-year BCR. The findings revealed that the radiomic score was associated with a distinctive tumor microenvironment.

The Cures Act Final Rule requires that patients have real-time access to their radiology reports, which contain technical language. Our objective to was to use a novel system called RadGPT, which integrates concept extraction and a large language model (LLM), to help patients understand their radiology reports. RadGPT generated 150 concept explanations and 390 question-and-answer pairs from 30 radiology report impressions from between 2012 and 2020. The extracted concepts were used to create concept-based explanations, as well as concept-based question-and-answer pairs where questions were generated using either a fixed template or an LLM. Additionally, report-based question-and-answer pairs were generated directly from the impression using an LLM without concept extraction. One board-certified radiologist and 4 radiology residents rated the material quality using a standardized rubric. Concept-based LLM-generated questions were significantly higher quality than concept-based template-generated questions (p < 0.001). Excluding those template-based question-and-answer pairs from further analysis, nearly all (> 95%) of RadGPT-generated materials were rated highly, with at least 50% receiving the highest possible ranking from all 5 raters. No answers or explanations were rated as likely to affect the safety or effectiveness of patient care. Report-level LLM-based questions and answers were rated particularly highly, with 92% of report-level LLM-based questions and 61% of the corresponding report-level answers receiving the highest rating from all raters. The educational tool RadGPT generated high-quality explanations and question-and-answer pairs that were personalized for each radiology report, unlikely to produce harmful explanations and likely to enhance patient understanding of radiology information.

Foundation models (FMs) are changing the way medical images are analyzed by learning from large collections of unlabeled data. Instead of relying on manually annotated examples, FMs are pre-trained to learn general-purpose visual features that can later be adapted to specific clinical tasks with little additional supervision. In this review, we examine how FMs are being developed and applied in pathology, radiology, and ophthalmology, drawing on evidence from over 150 studies. We explain the core components of FM pipelines, including model architectures, self-supervised learning methods, and strategies for downstream adaptation. We also review how FMs are being used in each imaging domain and compare design choices across applications. Finally, we discuss key challenges and open questions to guide future research.

The present study aims to construct a random forest (RF) model based on ultrasound radiomics and elastography, offering a new approach for the differentiation of thyroid nodules (TNs). We retrospectively analyzed 152 TNs from 127 patients and developed four machine learning models. The examination was performed using the Resona 9Pro equipped with a 15-4 MHz linear array probe. The region of interest (ROI) was delineated with 3D Slicer. Using the RF algorithm, four models were developed based on sound touch elastography (STE) parameters, strain elastography (SE) parameters, and the selected radiomic features: the STE model, SE model, radiomics model, and the combined model. Decision Curve Analysis (DCA) is employed to assess the clinical benefit of each model. The DeLong test is used to determine whether the area under the curves (AUC) values of different models are statistically significant. A total of 1396 radiomic features were extracted using the Pyradiomics package. After screening, a total of 7 radiomic features were ultimately included in the construction of the model. In STE, SE, radiomics model, and combined model, the AUCs are 0.699 (95% CI: 0.570-0.828), 0.812 (95% CI: 0.683-0.941), 0.851 (95% CI: 0.739-0.964) and 0.911 (95% CI: 0.806-1.000), respectively. In these models, the combined model and the radiomics model exhibited outstanding performance. The combined model, integrating elastography and radiomics, demonstrates superior predictive accuracy compared to single models, offering a promising approach for the diagnosis of TNs.

The integration of large language models (LLMs) into medical diagnostics has garnered substantial attention due to their potential to enhance diagnostic accuracy, streamline clinical workflows, and address health care disparities. However, the rapid evolution of LLM research necessitates a comprehensive synthesis of their applications, challenges, and future directions. This scoping review aimed to provide an overview of the current state of research regarding the use of LLMs in medical diagnostics. The study sought to answer four primary subquestions, as follows: (1) Which LLMs are commonly used? (2) How are LLMs assessed in diagnosis? (3) What is the current performance of LLMs in diagnosing diseases? (4) Which medical domains are investigating the application of LLMs? This scoping review was conducted according to the Joanna Briggs Institute Manual for Evidence Synthesis and adheres to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews). Relevant literature was searched from the Web of Science, PubMed, Embase, IEEE Xplore, and ACM Digital Library databases from 2022 to 2025. Articles were screened and selected based on predefined inclusion and exclusion criteria. Bibliometric analysis was performed using VOSviewer to identify major research clusters and trends. Data extraction included details on LLM types, application domains, and performance metrics. The field is rapidly expanding, with a surge in publications after 2023. GPT-4 and its variants dominated research (70/95, 74% of studies), followed by GPT-3.5 (34/95, 36%). Key applications included disease classification (text or image-based), medical question answering, and diagnostic content generation. LLMs demonstrated high accuracy in specialties like radiology, psychiatry, and neurology but exhibited biases in race, gender, and cost predictions. Ethical concerns, including privacy risks and model hallucination, alongside regulatory fragmentation, were critical barriers to clinical adoption. LLMs hold transformative potential for medical diagnostics but require rigorous validation, bias mitigation, and multimodal integration to address real-world complexities. Future research should prioritize explainable artificial intelligence frameworks, specialty-specific optimization, and international regulatory harmonization to ensure equitable and safe clinical deployment.

To evaluate the performance of machine learning (ML) models for the automated scoring of spinal MRI bone marrow oedema (BMO) in patients with axial spondyloarthritis (axSpA) and compare them with expert scoring. ML algorithms using SpineNet software were trained and validated on 3483 spinal MRIs from 686 axSpA patients across two clinical trial datasets. The scoring pipeline involved (i) detection and labelling of vertebral bodies and (ii) classification of vertebral units for the presence or absence of BMO. Two models were tested: Model 1, without manual segmentation, and Model 2, incorporating an intermediate manual segmentation step. Model outputs were compared with those of human experts using kappa statistics, balanced accuracy, sensitivity, specificity, and AUC. Both models performed comparably to expert readers, regarding presence vs absence of BMO. Model 1 outperformed Model 2, with an AUC of 0.94 (vs 0.88), accuracy of 75.8% (vs 70.5%), and kappa of 0.50 (vs 0.31), using absolute reader consensus scoring as the external reference; this performance was similar to the expert inter-reader accuracy of 76.8% and kappa of 0.47, in a radiographic axSpA dataset. In a non-radiographic axSpA dataset, Model 1 achieved an AUC of 0.97 (vs 0.91 for Model 2), accuracy of 74.6% (vs 70%), and kappa of 0.52 (vs 0.27), comparable to the expert inter-reader accuracy of 74.2% and kappa of 0.46. ML software shows potential for automated MRI BMO assessment in axSpA, offering benefits such as improved consistency, reduced labour costs, and minimised inter- and intra-reader variability. Clinicaltrials.gov, MEASURE 1 study (NCT01358175); PREVENT study (NCT02696031).

The accurate diagnosis of lymph node metastasis in esophageal squamous cell carcinoma is crucial in the treatment workflow, and the process is often time-consuming for clinicians. Recent deep learning models predicting whether lymph nodes are affected by cancer in esophageal cancer cases suffer from challenging node delineation and hence gain poor diagnosis accuracy. This paper proposes an innovative multi-task and multi-scale attention network (M <math xmlns="http://www.w3.org/1998/Math/MathML"><mmultiscripts><mrow></mrow> <mrow></mrow> <mn>2</mn></mmultiscripts> </math> ANet) to predict lymph node metastasis precisely. The network softly expands the regions of the node mask and subsequently utilizes the expanded mask to aggregate image features, thereby amplifying the node contexts. It additionally proposes a two-branch training strategy that compels the model to simultaneously predict metastasis probability and node masks, fostering a more comprehensive learning process. The node metastasis prediction performance has been evaluated on a self-collected dataset with 177 patients. Our model finally achieves a competitive accuracy of 83.7% on the test set comprising 577 nodes. With the adaptability to intricate patterns and ability to handle data variations, M <math xmlns="http://www.w3.org/1998/Math/MathML"><mmultiscripts><mrow></mrow> <mrow></mrow> <mn>2</mn></mmultiscripts> </math> ANet emerges as a promising tool for robust and comprehensive lymph node metastasis prediction in medical image analysis.

Alzheimer's disease (AD) is a heterogeneous neurodegenerative disorder in which tau neurofibrillary tangles are a pathological hallmark closely associated with cognitive dysfunction and neurodegeneration. In this study, we used brain tau data to investigate AD heterogeneity by identifying and characterizing the subpopulations among patients. We included 615 cognitively normal and 159 AD brain ¹⁸F-flortaucipr PET scans, along with T1-weighted MRI from the Alzheimer Disease Neuroimaging Initiative database. A three dimensional-convolutional neural network model was employed for AD detection using standardized uptake value ratio (SUVR) images. The model-derived saliency maps were generated and employed as informative image features for clustering AD participants. Among the identified subpopulations, statistical analysis of demographics, neuropsychological measures, and SUVR were compared. Correlations between neuropsychological measures and regional SUVRs were assessed. A generalized linear model was utilized to investigate the sex and APOE ε4 interaction effect on regional SUVRs. Two distinct subpopulations of AD patients were revealed, denoted as S_Hi and S_Lo. Compared to the S_Lo group, the S_Hi group exhibited a significantly higher global tau burden in the brain, but both groups showed similar cognition distribution levels. In the S_Hi group, the associations between the neuropsychological measurements and regional tau deposition were weakened. Moreover, a significant interaction effect of sex and APOE ε4 on tau deposition was observed in the S_Lo group, but no such effect was found in the S_Hi group. Our results suggest that tau tangles, as shown by SUVR, continue to accumulate even when cognitive function plateaus in AD patients, highlighting the advantages of PET in later disease stages. The differing relationships between cognition and tau deposition, and between gender, APOE4, and tau deposition, provide potential for subtype-specific treatments. Targeting gender-specific and genetic factors influencing tau deposition, as well as interventions aimed at tau's impact on cognition, may be effective.

Missed fractures are the primary cause of interpretation errors in emergency radiology, and artificial intelligence has recently shown great promise in radiograph interpretation. This study compared the diagnostic performance of two AI algorithms, BoneView and RBfracture, in detecting traumatic abnormalities (fractures and dislocations) in MSK radiographs. AI algorithms analyzed 998 radiographs (585 normal, 413 abnormal), against the consensus of two MSK specialists. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and interobserver agreement (Cohen's Kappa) were calculated. 95% confidence intervals (CI) assessed robustness, and McNemar's tests compared sensitivity and specificity between the AI algorithms. BoneView demonstrated a sensitivity of 0.893 (95% CI: 0.860-0.920), specificity of 0.885 (95% CI: 0.857-0.909), PPV of 0.846, NPV of 0.922, and accuracy of 0.889. RBfracture demonstrated a sensitivity of 0.872 (95% CI: 0.836-0.901), specificity of 0.892 (95% CI: 0.865-0.915), PPV of 0.851, NPV of 0.908, and accuracy of 0.884. No statistically significant differences were found in sensitivity (p = 0.151) or specificity (p = 0.708). Kappa was 0.81 (95% CI: 0.77-0.84), indicating almost perfect agreement between the two AI algorithms. Performance was similar in adults and children. Both AI algorithms struggled more with subtle abnormalities, which constituted 66% and 70% of false negatives but only 20% and 18% of true positives for the two AI algorithms, respectively (p < 0.001). BoneView and RBfracture exhibited high diagnostic performance and almost perfect agreement, with consistent results across adults and children, highlighting the potential of AI in emergency radiograph interpretation.

Brain aneurysm detection models, both in the literature and in industry, continue to lack generalizability during external validation, limiting clinical adoption. This challenge is largely due to extensive exclusion criteria during training data selection. The authors developed the first model to achieve generalizability using novel methodological approaches. Computed tomography angiography (CTA) scans from 2004 to 2023 at the study institution were used for model training, including untreated unruptured intracranial aneurysms without extensive cerebrovascular disease. External validation used digital subtraction angiography-verified CTAs from an international center, while prospective validation occurred at the internal institution over 9 months. A public web platform was created for further model validation. A total of 2194 CTA scans were used for this study. One thousand five hundred eighty-seven patients and 1920 aneurysms with a mean size of 5.3 ± 3.7 mm were included in the training cohort. The mean age of the patients was 69.7 ± 14.9 years, and 1203 (75.8%) were female. The model achieved a training Dice score of 0.88 and a validation Dice score of 0.76. Prospective internal validation on 304 scans yielded a lesion-level (LL) sensitivity of 82.5% (95% CI: 75.5-87.9) and specificity of 89.6 (95% CI: 84.5-93.2). External validation on 303 scans demonstrated an on-par LL sensitivity and specificity of 83.5% (95% CI: 75.1-89.4) and 92.9% (95% CI: 88.8-95.6), respectively. Radiologist LL sensitivity from the external center was 84.5% (95% CI: 76.2-90.2), and 87.5% of the missed aneurysms were detected by the model. The authors developed the first publicly testable artificial intelligence model for aneurysm detection on CTA scans, demonstrating generalizability and state-of-the-art performance in external validation. The model addresses key limitations of previous efforts and enables broader validation through a web-based platform.