To explore the value of machine learning models based on MRI radiomics and automated habitat analysis in predicting bone metastasis and high-grade pathological Gleason scores in prostate cancer. This retrospective study enrolled 214 patients with pathologically diagnosed prostate cancer from May 2013 to January 2025, including 93 cases with bone metastasis and 159 cases with high-grade Gleason scores. Clinical, pathological and MRI data were collected. An nnUNet model automatically segmented the prostate in MRI scans. K-means clustering identified subregions within the entire prostate in T2-FS images. Senior radiologists manually segmented regions of interest (ROIs) in prostate lesions. Radiomics features were extracted from these habitat subregions and lesion ROIs. These features combined with clinical features were utilized to build multiple machine learning classifiers to predict bone metastasis and high-grade Gleason scores while a K-means clustering method was applied to obtain habitat subregions within the whole prostate. Finally, the models underwent interpretable analysis based on feature importance. The nnUNet model achieved a mean Dice coefficient of 0.970 for segmentation. Habitat analysis using 2 clusters yielded the highest average silhouette coefficient (0.57). Machine learning models based on a combination of lesion radiomics, habitat radiomics, and clinical features achieved the best performance in both prediction tasks. The Extra Trees Classifier achieved the highest AUC (0.900) for predicting bone metastasis, while the CatBoost Classifier performed best (AUC 0.895) for predicting high-grade Gleason scores. The interpretability analysis of the optimal models showed that the PSA clinical feature was crucial for predictions, while both habitat radiomics and lesion radiomics also played important roles. The study proposed an automated prostate habitat analysis for prostate cancer, enabling a comprehensive analysis of tumor heterogeneity. The machine learning models developed achieved excellent performance in predicting the risk of bone metastasis and high-grade Gleason scores in prostate cancer. This approach overcomes the limitations of manual feature extraction, and the inadequate analysis of heterogeneity often encountered in traditional radiomics, thereby improving model performance.

PURPOSECardiac amyloidosis (CA), a life-threatening infiltrative cardiomyopathy, can be non-invasively diagnosed using [99mTc]Tc-bisphosphonate SPECT/CT. However, subjective visual interpretation risks diagnostic inaccuracies. We developed and validated a machine learning (ML) framework leveraging SPECT/CT radiomics to automate CA detection. METHODSThis retrospective multicenter study analyzed 290 patients of suspected CA who underwent [99mTc]Tc-PYP or [99mTc]Tc-DPD SPECT/CT. Radiomic features were extracted from co-registered SPECT and CT images, harmonized via intra-class correlation and Pearson correlation filtering, and optimized through LASSO regression. A stacking ensemble model incorporating support vector machine (SVM), random forest (RF), gradient boosting decision tree (GBDT), and adaptive boosting (AdaBoost) classifiers was constructed. The model was validated using an internal validation set (n = 54) and two external test set (n = 54 and n = 58).Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis (DCA). Feature importance was interpreted using SHapley Additive exPlanations (SHAP) values. RESULTSOf 290 patients, 117 (40.3%) had CA. The stacking radiomics model attained AUCs of 0.871, 0.824, and 0.839 in the validation, test 1, and test 2 cohorts, respectively, significantly outperforming the clinical model (AUC 0.546 in validation set, P<0.05). DCA demonstrated superior net benefit over the clinical model across relevant thresholds, and SHAP analysis highlighted wavelet-transformed first-order and texture features as key predictors. CONCLUSIONA stacking ML model with SPECT/CT radiomics improves CA diagnosis, showing strong generalizability across varied imaging protocols and populations and highlighting its potential as a decision-support tool.

Coronary computed tomography angiography (CCTA) is a first-line investigation for chest pain in patients with suspected obstructive coronary artery disease (CAD). However, many acute cardiac events occur in the absence of obstructive CAD. We assessed the lifetime cost-effectiveness of integrating a novel artificial intelligence-enhanced image analysis algorithm (AI-Risk) that stratifies the risk of cardiac events by quantifying coronary inflammation, combined with the extent of coronary artery plaque and clinical risk factors, by analysing images from routine CCTA. A hybrid decision-tree with population cohort Markov model was developed from 3393 consecutive patients who underwent routine CCTA for suspected obstructive CAD and followed up for major adverse cardiac events over a median (interquartile range) of 7.7(6.4-9.1) years. In a prospective real-world evaluation survey of 744 consecutive patients undergoing CCTA for chest pain investigation, the availability of AI-Risk assessment led to treatment initiation or intensification in 45% of patients. In a further prospective study of 1214 consecutive patients with extensive guidelines recommended cardiovascular risk profiling, AI-Risk stratification led to treatment initiation or intensification in 39% of patients beyond the current clinical guideline recommendations. Treatment guided by AI-Risk modelled over a lifetime horizon could lead to fewer cardiac events (relative reductions of 11%, 4%, 4%, and 12% for myocardial infarction, ischaemic stroke, heart failure, and cardiac death, respectively). Implementing AI-Risk Classification in routine interpretation of CCTA is highly likely to be cost-effective (incremental cost-effectiveness ratio £1371-3244), both in scenarios of current guideline compliance, or when applied only to patients without obstructive CAD. Compared with standard care, the addition of AI-Risk assessment in routine CCTA interpretation is cost-effective, by refining risk-guided medical management.

This study aimed to evaluate the ability of ChatGPT and Breast Ultrasound Helper, a special ChatGPT-based subprogram trained on ultrasound image analysis, to analyze and differentiate benign and malignant breast lesions on ultrasound images. Ultrasound images of histopathologically confirmed breast cancer and fibroadenoma patients were read GPT-4o (the latest ChatGPT version) and Breast Ultrasound Helper (BUH), a tool from the "Explore" section of ChatGPT. Both were prompted in English using ACR BI-RADS Breast Ultrasound Lexicon criteria: lesion shape, orientation, margin, internal echo pattern, echogenicity, posterior acoustic features, microcalcifications or hyperechoic foci, perilesional hyperechoic rim, edema or architectural distortion, lesion size, and BI-RADS category. Two experienced radiologists evaluated the images and the responses of the programs in consensus. The outputs, BI-RADS category agreement, and benign/malignant discrimination were statistically compared. A total of 232 ultrasound images were analyzed, of which 133 (57.3%) were malignant and 99 (42.7%) benign. In comparative analysis, BUH showed superior performance overall, with higher kappa values and statistically significant results across multiple features (P .001). However, the overall level of agreement with the radiologists' consensus for all features was similar for BUH (κ: 0.387-0.755) and GPT-4o (κ: 0.317-0.803). On the other hand, BI-RADS category agreement was slightly higher in GPT-4o than in BUH (69.4% versus 65.9%), but BUH was slightly more successful in distinguishing benign lesions from malignant lesions (65.9% versus 67.7%). Although both AI tools show moderate-good performance in ultrasound image analysis, their limited compatibility with radiologists' evaluations and BI-RADS categorization suggests that their clinical application in breast ultrasound interpretation is still early and unreliable.

This study employed a retrospective data analysis approach combined with model development and validation. The present study introduces a 2.5D convolutional neural network (CNN) model leveraging CT imaging to facilitate the early detection of malignant vertebral compression fractures (MVCFs), potentially reducing reliance on invasive biopsies. Vertebral histopathological biopsy is recognized as the gold standard for differentiating between osteoporotic and malignant vertebral compression fractures (VCFs). Nevertheless, its application is restricted due to its invasive nature and high cost, highlighting the necessity for alternative methods to identify MVCFs. The clinical, imaging, and pathological data of patients who underwent vertebral augmentation and biopsy at Institution 1 and Institution 2 were collected and analyzed. Based on the vertebral CT images of these patients, 2D, 2.5D, and 3D CNN models were developed to identify the patients with osteoporotic vertebral compression fractures (OVCF) and MVCF. To verify the clinical application value of the CNN model, two rounds of reader studies were performed. The 2.5D CNN model performed well, and its performance in identifying MVCF patients was significantly superior to that of the 2D and 3D CNN models. In the training dataset, the area under the receiver operating characteristic curve (AUC) of the 2.5D CNN model was 0.996 and an F1 score of 0.915. In the external cohort test, the AUC was 0.815 and an F1 score of 0.714. And clinicians' ability to identify MVCF patients has been enhanced by the 2.5D CNN model. With the assistance of the 2.5D CNN model, the AUC of senior clinicians was 0.882, and the F1 score was 0.774. For junior clinicians, the 2.5D CNN model-assisted AUC was 0.784 and the F1 score was 0.667. The development of our 2.5D CNN model marks a significant step towards non-invasive identification of MVCF patients,. The 2.5D CNN model may be a potential model to assist clinicians in better identifying MVCF patients.

Extracting clinical entities from unstructured medical documents is critical for improving clinical decision support and documentation workflows. This study examines the performance of various encoder and decoder models trained for Named Entity Recognition (NER) of clinical parameters in pathology and radiology reports, highlighting the applicability of Large Language Models (LLMs) for this task. Three NER methods were evaluated: (1) flat NER using transformer-based models, (2) nested NER with a multi-task learning setup, and (3) instruction-based NER utilizing LLMs. A dataset of 2013 pathology reports and 413 radiology reports, annotated by medical students, was used for training and testing. The performance of encoder-based NER models (flat and nested) was superior to that of LLM-based approaches. The best-performing flat NER models achieved F1-scores of 0.87-0.88 on pathology reports and up to 0.78 on radiology reports, while nested NER models performed slightly lower. In contrast, multiple LLMs, despite achieving high precision, yielded significantly lower F1-scores (ranging from 0.18 to 0.30) due to poor recall. A contributing factor appears to be that these LLMs produce fewer but more accurate entities, suggesting they become overly conservative when generating outputs. LLMs in their current form are unsuitable for comprehensive entity extraction tasks in clinical domains, particularly when faced with a high number of entity types per document, though instructing them to return more entities in subsequent refinements may improve recall. Additionally, their computational overhead does not provide proportional performance gains. Encoder-based NER models, particularly those pre-trained on biomedical data, remain the preferred choice for extracting information from unstructured medical documents.

Pathological complete response (pCR) to neoadjuvant systemic therapy (NAST) is an established prognostic marker in breast cancer (BC). Multimodal deep learning (DL), integrating diverse data sources (radiology, pathology, omics, clinical), holds promise for improving pCR prediction accuracy. This systematic review synthesizes evidence on multimodal DL for pCR prediction and compares its performance against unimodal DL. Following PRISMA, we searched PubMed, Embase, and Web of Science (January 2015-April 2025) for studies applying DL to predict pCR in BC patients receiving NAST, using data from radiology, digital pathology (DP), multi-omics, and/or clinical records, and reporting AUC. Data on study design, DL architectures, and performance (AUC) were extracted. A narrative synthesis was conducted due to heterogeneity. Fifty-one studies, mostly retrospective (90.2%, median cohort 281), were included. Magnetic resonance imaging and DP were common primary modalities. Multimodal approaches were used in 52.9% of studies, often combining imaging with clinical data. Convolutional neural networks were the dominant architecture (88.2%). Longitudinal imaging improved prediction over baseline-only (median AUC 0.91 vs. 0.82). Overall, the median AUC across studies was 0.88, with 35.3% achieving AUC ≥ 0.90. Multimodal models showed a modest but consistent improvement over unimodal approaches (median AUC 0.88 vs. 0.83). Omics and clinical text were rarely primary DL inputs. DL models demonstrate promising accuracy for pCR prediction, especially when integrating multiple modalities and longitudinal imaging. However, significant methodological heterogeneity, reliance on retrospective data, and limited external validation hinder clinical translation. Future research should prioritize prospective validation, integration underutilized data (multi-omics, clinical), and explainable AI to advance DL predictors to the clinical setting.

Endoscopic image classification plays a pivotal role in medical diagnostics by identifying anatomical landmarks and pathological findings. However, conventional closed-set classification frameworks are inherently limited in open-world clinical settings, where previously unseen conditions can arise andcompromise model reliability. To address this, we explore the application of Open Set Recognition (OSR) techniques on the Kvasir dataset, a publicly available and diverse endoscopic image collection. In this study, we evaluate and compare the OSR capabilities of several representative deep learning architectures, including ResNet-50, Swin Transformer, and a hybrid ResNet-Transformer model, under both closed-set and open-set conditions. OpenMax is adopted as a baseline OSR method to assess the ability of these models to distinguish known classes from previously unseen categories. This work represents one of the first efforts to apply open set recognition to the Kvasir dataset and provides a foundational benchmark for evaluating OSR performance in medical image analysis. Our results offer practical insights into model behavior in clinically realistic settings and highlight the importance of OSR techniques for the safe deployment of AI systems in endoscopy.

Chest X-ray (CXR) is the most frequently ordered imaging test, supporting diverse clinical tasks from thoracic disease detection to postoperative monitoring. However, task-specific classification models are limited in scope, require costly labeled data, and lack generalizability to out-of-distribution datasets. To address these challenges, we introduce CheXFound, a self-supervised vision foundation model that learns robust CXR representations and generalizes effectively across a wide range of downstream tasks. We pretrained CheXFound on a curated CXR-987K dataset, comprising over approximately 987K unique CXRs from 12 publicly available sources. We propose a Global and Local Representations Integration (GLoRI) head for downstream adaptations, by incorporating fine- and coarse-grained disease-specific local features with global image features for enhanced performance in multilabel classification. Our experimental results showed that CheXFound outperformed state-of-the-art models in classifying 40 disease findings across different prevalence levels on the CXR-LT 24 dataset and exhibited superior label efficiency on downstream tasks with limited training data. Additionally, CheXFound achieved significant improvements on downstream tasks with out-of-distribution datasets, including opportunistic cardiovascular disease risk estimation, mortality prediction, malpositioned tube detection, and anatomical structure segmentation. The above results demonstrate CheXFound's strong generalization capabilities, which will enable diverse downstream adaptations with improved label efficiency in future applications. The project source code is publicly available at https://github.com/RPIDIAL/CheXFound.

Stationary functional brain networks (sFBNs) and dynamic functional brain networks (dFBNs) derived from resting-state functional MRI characterize the complex interactions of the human brain from different aspects and could offer complementary information for brain disease analysis. Most current studies focus on sFBN or dFBN analysis, thus limiting the performance of brain network analysis. A few works have explored integrating sFBN and dFBN to identify brain diseases, and achieved better performance than conventional methods. However, these studies still ignore some valuable discriminative information, such as the distribution information of subjects between and within categories. This paper presents a Double Collaborative Learning Network (DCLNet), which takes advantage of both collaborative encoder and collaborative contrastive learning, to learn complementary information of sFBN and dFBN and distribution information of subjects between inter- and intra-categories for brain disease classification. Specifically, we first construct sFBN and dFBN using traditional correlation-based methods with rs-fMRI data, respectively. Then, we build a collaborative encoder to extract brain network features at different levels (i.e., connectivity-based, brain-region-based, and brain-network-based features), and design a prune-graft transformer module to embed the complementary information of the features at each level between two kinds of FBNs. We also develop a collaborative contrastive learning module to capture the distribution information of subjects between and within different categories, thereby learning the more discriminative features of brain networks. We evaluate the DCLNet on two real brain disease datasets with rs-fMRI data, with experimental results demonstrating the superiority of the proposed method.