Informatics in Medicine Unlocked最新文献

{"title":"When does retrieval help medical question answering? A question-type sensitivity analysis and failure taxonomy for text-based (image-free) medical RAG, with an interpretable clinical filter (CAF-RAG)","authors":"Youssef Azami,&nbsp;Hamza Boulibya,&nbsp;Mohamed Adnane Mahraz,&nbsp;Mohammed Oulhadj,&nbsp;Hamid Tairi,&nbsp;Ilyasse Aboussaleh,&nbsp;Khalid El Fazazy,&nbsp;Jamal Riffi","doi":"10.1016/j.imu.2026.101810","DOIUrl":"10.1016/j.imu.2026.101810","url":null,"abstract":"<div><div>Retrieval-augmented generation (RAG) is widely assumed to improve medical question answering, yet the conditions under which retrieval helps, leaves performance unchanged, or actively harms it remain poorly characterized. This paper is a diagnostic study of that question rather than the proposal of a new retrieval framework. We evaluate a deliberately text-only (image-free) pipeline in which the model receives the question and retrieved text but never the medical image, using SLAKE and VQA-RAD as sources of clinically diverse question types, and we validate externally on PubMedQA (<span><math><mrow><mi>n</mi><mo>=</mo><mn>1000</mn></mrow></math></span>). We compare six retrieval conditions across two language models. The conditions are parametric-only No-RAG, dense RAG, hybrid BM25<!--> <!-->+<!--> <!-->dense, two cross-encoder rerankers (general and biomedical), and CAF-RAG, a training-free clinical filter whose auditable keep/remove decisions make it a diagnostic instrument. The two models are GPT-4o-mini and the open medical model BioMistral-7B, which read byte-identical retrieved context. Our central finding is negative and consistent: text-based retrieval does not significantly outperform parametric knowledge on any group we tested, and on external data it is significantly worse (PubMedQA, −0.120, <span><math><mrow><mi>p</mi><mo>&lt;</mo><mn>0.0001</mn></mrow></math></span>). No passage-selection method recovers it: neither reranker significantly beats plain RAG, and CAF-RAG is statistically indistinguishable from a cross-encoder reranker (＋0.002, <span><math><mrow><mi>p</mi><mo>=</mo><mn>0.91</mn></mrow></math></span>) while recovering part of the accuracy that retrieval costs (＋0.034 over RAG on SLAKE-CLOSED, <span><math><mrow><mi>p</mi><mo>=</mo><mn>0.027</mn></mrow></math></span>). The two models agree to within 0.002 on both effects, indicating these are properties of retrieval rather than of one model. We identify a behavioral mechanism: retrieved context displaces the model from its parametric answer prior. It moves the model toward affirmation when the context is topically related to the question (both models swing sharply toward “Yes” on SLAKE), and toward hedging when it is not (both swing toward “Maybe” on PubMedQA). We also report an audit of our own evaluation: a substring-matching artifact in lexical scoring inflated retrieval conditions more than the terse parametric baseline, and we replace it with a scorer validated against 197 blinded human gradings (91.9% agreement, <span><math><mrow><mi>κ</mi><mo>=</mo><mn>0.79</mn></mrow></math></span>). All per-question outputs, the rebuilt knowledge base, retrieval contexts, and the scorer with its documented failure modes are released.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"65 ","pages":"Article 101810"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148952497","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Factorial evaluation of DICOM harmonization and clinically constrained augmentation for multi-vendor TB triage","authors":"Abu Salam,&nbsp;Ika Novita Dewi,&nbsp;Pulung Nurtantio Andono,&nbsp;Purwanto,&nbsp;Moch Arief Soeleman,&nbsp;Catur Supriyanto,&nbsp;Ardytha Luthfiarta,&nbsp;Eko Adhi Pangarsa,&nbsp;Lilik Lestari,&nbsp;Eric Araka","doi":"10.1016/j.imu.2026.101808","DOIUrl":"10.1016/j.imu.2026.101808","url":null,"abstract":"<div><div>Tuberculosis (TB) triage from chest radiographs is influenced by image representation and training-data variability in addition to model architecture. This study quantified the independent and interaction effects of these workflow factors using 3522 examinations from 3100 patients in a single-institution multi-vendor PACS/DICOM cohort. A representation-matched 2 × 2 factorial design compared a conventional DICOM-to-PNG representation (P0) with DICOM-aware HyFusion preprocessing (P1), and the original training set (A0) with a deterministic clinically constrained augmented pool (A1). Four DenseNet121 conditions used identical patient-level partitions, optimization settings, three random seeds, and a fixed raw-probability threshold of 0.5. Evaluation included discrimination, operational burden per 1000 examinations, validation-only temperature scaling, 5000 paired patient-level bootstrap resamples, and adjusted generalized estimating equations. C3 (P1A0) achieved the highest mean AUROC (0.8732 ± 0.0085) and AUPRC (0.8809 ± 0.0194). Relative to C1 (P0A0), the AUROC and AUPRC differences were ＋0.0185 (95% CI −0.0070 to ＋0.0451) and ＋0.0217 (95% CI −0.0023 to ＋0.0487), respectively. At the fixed threshold, C3 increased specificity and PPV and reduced referrals by 78.5 and false positives by 53.5 per 1000 examinations. A1 reduced referrals and false positives but also reduced sensitivity and increased false negatives. Temperature scaling improved NLL, Brier score, and ECE10 across all conditions. The study establishes a reproducible workflow-factor attribution framework and identifies P1A0 as the DenseNet121 configuration with the highest mean discrimination and a favorable calibration and referral-efficiency profile in this cohort.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"65 ","pages":"Article 101808"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148952496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Retrospective machine learning approach for forecasting in-hospital death in ICU patients after cardiac arrest","authors":"Yong Si,&nbsp;Li Sun,&nbsp;Shuheng Chen,&nbsp;JunYi Fan,&nbsp;Elham Pishgar,&nbsp;Kamiar Alaei,&nbsp;Greg Placencia,&nbsp;Maryam Pishgar","doi":"10.1016/j.imu.2026.101776","DOIUrl":"10.1016/j.imu.2026.101776","url":null,"abstract":"<div><div>Accurate identification of patients at high risk of in-hospital mortality in intensive care units (ICUs) is vital for enhancing clinical decision-making and improving patient care strategies. As traditional statistical models often fall short in modeling nonlinear and multifactorial clinical variables, this study explores a machine learning (ML) approach to overcome these limitations.</div><div>We conducted a retrospective study using the MIMIC-IV database, focusing on 1132 adult ICU patients with cardiac arrest identified from index ICU admissions after applying predefined inclusion and exclusion criteria. Numerical features were summarized using mean aggregation over the first 48 h of ICU admission to provide a compact and clinically interpretable early-risk representation, while categorical attributes underwent structured encoding. The dataset was split into 70% for training and 30% for testing. We applied a combination of regularization techniques (LASSO, Ridge, ElasticNet) and Random Forest-based importance ranking for feature selection.</div><div>Multiple supervised ML algorithms, including Logistic Regression, LightGBM, CatBoost, XGBoost, and a feed-forward Neural Network, were benchmarked using metrics such as AUC-ROC, calibration plots, and decision curve analysis.</div><div>The XGBoost algorithm achieved the most favorable results with a test AUC of 0.804 (95% CI: 0.760–0.849) and accuracy of 0.762. Key predictors identified by the final model included neurological status measures, lactate, sodium, anion gap, and coagulation-related indices. These findings suggest that the proposed model offers a reliable, interpretable, and potentially deployable framework for ICU mortality risk prediction.</div><div>Unlike prior MIMIC-based mortality studies that primarily focused on discrimination, our framework additionally emphasizes calibration, decision-analytic utility, and model interpretability for early ICU risk stratification. Rather than replacing traditional ICU severity scores, the proposed framework is intended to augment them through integrated risk modeling and interpretable prediction.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101776"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148554997","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Beyond one stain: Cross-Modality feature learning boosts accuracy in HER2 status scoring","authors":"Olaide N. Oyelade,&nbsp;Oliver Hoxey,&nbsp;Yulia Humrye","doi":"10.1016/j.imu.2026.101785","DOIUrl":"10.1016/j.imu.2026.101785","url":null,"abstract":"<div><div>The widespread use of histopathology images, particularly those stained with hematoxylin and eosin (H&amp;E), has proven pivotal for detecting tumors. However, advancing cancer cases to appropriate treatment pathways requires accurate quantification of human epidermal growth factor receptor 2 (HER2) protein expression. Predicting both low- and high-level expressions of HER2 remains challenging. Moreover, the joint analysis of H&amp;E and immunohistochemistry (IHC) stained images for HER2 scoring is difficult. Although several deep learning methods have been investigated to address this clinical task, they fail to provide pixel-level localization of HER2 status. In this study, we propose a single end-to-end pipeline using a system of vision transformers with HER2 status scoring on whole slide images of WSIs. The method includes patch-wise processing of H&amp;E WSIs for tumor localization. A novel mapping function is proposed to correspondingly identify correlated IHC WSI regions with malignant regions on H&amp;E. A clinically inspired HER2 scoring mechanism is embedded in the pipeline and allows for automatic pixel-level annotation of 4-way HER2 scoring (0, 1+, 2+, and 3+). Also, the proposed method accurately returns HER2-negative and HER2-positive. Privately curated datasets were collaboratively extracted from 13 different cases of WSIs of H&amp;E and IHC. A thorough experiment was conducted on the proposed method. Results obtained showed good classification accuracy during tumor localization. Also, a classification accuracy of 0.94 and a specificity of 0.933 were obtained for the prediction of HER2 status, scoring in the 4-way methods. The applicability of the proposed pipeline was investigated using WSIs patches as comparable to human pathologists. Findings from the study showed the usability of jointly evaluated H&amp;E and IHC images on end-to-end ViT-based models for HER2 scoring.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101785"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148554998","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning workflow for 3D calcification segmentation in cone beam computed tomographic images","authors":"Kimia Darvish Noori,&nbsp;Hamed Bavarnegin,&nbsp;Kyle Reeves,&nbsp;Seyed Abolghasem Mirroshandel,&nbsp;Mina Mahdian","doi":"10.1016/j.imu.2026.101784","DOIUrl":"10.1016/j.imu.2026.101784","url":null,"abstract":"<div><div>Calcium salt deposits, commonly referred to as calcifications, can form in various body tissues. Among them, calcifications in the cervical carotid arteries are of particular clinical significance due to their strong association with cardiovascular disease. These calcifications are often detected as incidental findings, that is, unexpected observations that may carry greater clinical importance than the original purpose of the imaging. Given their diagnostic relevance, automated detection and segmentation of such calcifications in medical images are critical for enabling early diagnosis and improving clinical decision-making. Despite their importance, detecting carotid artery calcifications presents several challenges: their small size, sparse distribution, and visual similarity to surrounding anatomical structures make them difficult to identify accurately. To address these challenges, this study proposes a multi-stage workflow for segmenting carotid artery calcifications. First, we introduce a novel 3D dataset comprising Cone Beam Computed Tomography (CBCT) scans that contain incidental neck calcifications. For the segmentation task, we adapt 3D UNETR, a vision transformer-based model. We then employ a 3D Grad-CAM technique to generate class activation maps, which are subsequently used in a post-processing step to refine the segmentation results. The proposed workflow achieves a Dice coefficient of 64% for segmentation, despite calcifications occupying only 0.04% of the total image volume. Beyond its strong performance on such sparse targets, the approach remains computationally efficient, with modest memory requirements and fast inference times, making it practical for integration into clinical imaging workflows.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101784"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148555005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From ensembles to knowledge-augmented reasoning: Information fusion for depression detection in the generative AI era","authors":"Zeynab Davoudi,&nbsp;Meysam Roostaee,&nbsp;Khadijeh Aghajani","doi":"10.1016/j.imu.2026.101781","DOIUrl":"10.1016/j.imu.2026.101781","url":null,"abstract":"<div><div>Computational depression detection has reached a saturation point with standard end-to-end deep learning paradigms. To navigate the rapid proliferation of highly complex models, this systematic scoping review analyzes 87 primary post-LLM studies, spanning both “in-the-wild” social media and structured clinical interactions, published primarily between January 2024 and April 2026. Evaluated alongside 7 foundational 2023 baselines (totaling 94 core methodology papers out of 111 cited references), these works mark a distinct methodological transition from isolated feature engineering to multimodal and architectural synthesis. We introduce a novel taxonomy that categorizes these emerging approaches into four information fusion paradigms: (1) Architecture Fusion, which hybridizes heterogeneous neural components to capture complementary spatiotemporal dependencies; (2) Knowledge Fusion, which injects structured external clinical knowledge (e.g., DSM-5 ontologies) into neural pipelines to bridge the semantic gap; (3) Generative Fusion, which integrates in-context learning (ICL), retrieval-augmented generation (RAG), and parameter-efficient fine-tuning (PEFT); and (4) Data-Level Fusion, which addresses multimodal alignment, dynamic fusion timing, and temporal modeling. Our critical analysis reveals that while zero-shot large language models (LLMs) offer broad semantic reasoning capabilities, fine-tuned encoders and hybrid architectures integrated with domain-specific knowledge currently provide a more rigorously validated balance of reliability, explainability, and performance. We conclude by exposing the “illusion of accuracy” prevalent in proxy-labeled benchmarks and outlining strategic directions for developing clinically valid, privacy-preserving detection systems suitable for real-world clinical workflows.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101781"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148555000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Manifold learning for personalized and label-free detection of cardiac arrhythmias","authors":"Amir Reza Vazifeh,&nbsp;Jason W. Fleischer","doi":"10.1016/j.imu.2026.101770","DOIUrl":"10.1016/j.imu.2026.101770","url":null,"abstract":"<div><div>Electrocardiograms (ECGs) provide non-invasive measurements of heart activity and are well-established tools for detecting cardiac arrhythmias. Although supervised machine learning has emerged as a promising approach for automated heartbeat classification, substantial variations in ECG signals across individuals and leads, combined with inconsistent labeling standards and dataset biases, make it difficult to develop generalizable models. Dimensionality reduction is an unsupervised approach that maps high-dimensional data into a lower-dimensional space while preserving the underlying structure, enabling data visualization and pattern discovery. Conventional methods, e.g., principal component analysis, prioritize large (and often obvious) variances in the data and typically overlook subtle yet clinically relevant patterns. Here, we show that nonlinear dimensionality reduction (NLDR) algorithms, e.g., t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP), can identify medically relevant features in ECG signals without pretraining or prior information. Using the MIT-BIH Arrhythmia Database, we show that: a) applying NLDR to a mixed population of heartbeats reveals inter-individual differences in morphology, as signals from the same person cluster together in latent spaces; and b) applying NLDR to the heartbeats of a single individual separates normal beats from arrhythmias into distinct clusters, which can be identified by clustering algorithms in an unsupervised manner. To our knowledge, this is the first systematic evaluation of NLDR for unsupervised arrhythmia detection. As a proof-of-concept, we demonstrate that both UMAP and t- SNE achieved trustworthiness scores of ≥0.95, indicating that local neighborhoods from the original space are well preserved in the embedding. Moreover, classification on the 2D embeddings outperforms classification in the original high-dimensional space, with a simple k-NN classifier discriminating individual recordings with ≥80% accuracy and identifying arrhythmias with a median accuracy of ≥98% and median F1-score of ≥85%. These results show that NLDR holds much promise for cardiac monitoring, and for personalized healthcare beyond cardiology.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101770"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148554615","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gait prediction post-botulinum toxin injection: A transfer learning approach","authors":"Michel Le Dez,&nbsp;Mathieu Lempereur,&nbsp;Maxime Devanne","doi":"10.1016/j.imu.2026.101778","DOIUrl":"10.1016/j.imu.2026.101778","url":null,"abstract":"<div><div>Accurate prediction of post-treatment gait outcomes in children with cerebral palsy (CP) remains a major challenge in clinical practice. Deep learning offers promising opportunities to model complex kinematic changes, particularly following botulinum toxin (BT) injections.</div><div>This study aimed to evaluate the ability of two neural network architectures, a bidirectional long short-term memory (Bi-LSTM) and a one-dimensional convolutional neural network (1D CNN), to predict post-BT gait kinematics from pre-treatment data.</div><div>Gait data from children with CP were analysed using both models, trained either from scratch or with transfer learning. Performance was assessed using root mean square error (RMSE) across joint angles, planes of motion, and individual subjects. Statistical comparisons were performed with Wilcoxon tests.</div><div>Bi-LSTM and 1D CNN models achieved comparable levels of accuracy, with average root mean square error (RMSE) values around 8-9°. Transfer learning significantly improved model performance across error- and correlation-based metrics, with especially pronounced improvements for 1D CNN model (RMSE=9.81°from scratch ; RMSE=7.39°with fine-tuning) in clinically relevant joint movements such as hip and knee flexion/extension.</div><div>These findings highlight the potential of deep learning models not only to capture complex gait patterns but also to adapt to treatment-induced changes. These models could support treatment planning, improve patient counselling, and contribute to the development of personalized rehabilitation strategies. Future work should expand to larger, multicenter datasets and emphasize clinical interpretability to facilitate translation into practice.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101778"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148554616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The prediction of all-cause mortality in end-stage kidney disease patients using social determinants of health: A machine learning framework","authors":"Addy Smith,&nbsp;Maxwell Donelan,&nbsp;Hossein Moradi Rekabdarkolaee,&nbsp;Patti Brooks,&nbsp;Brandon M. Varilek,&nbsp;Surachat Ngorsuraches,&nbsp;Jerry Schrier,&nbsp;Adam Dell,&nbsp;Semhar Michael","doi":"10.1016/j.imu.2026.101779","DOIUrl":"10.1016/j.imu.2026.101779","url":null,"abstract":"<div><div>End-stage kidney disease (ESKD) is the irreversible final stage of chronic kidney disease in which the kidneys lose their independent function. This study presents a machine learning framework to predict all-cause mortality in ESKD patients by the end of a follow-up period. We combined patient-specific clinical factors with social determinants of health (SDOH) to assess their influence on survival outcomes. Data were obtained from the United States Renal Data System, including patients admitted in 2015 and followed through August 2021. Community-level SDOH data were integrated from the Agency for Healthcare Research and Quality, with variable screening techniques and expert input guiding feature selection. To address class imbalance, the synthetic minority oversampling technique (SMOTE) was applied. Three machine learning models were developed: logistic regression, random forest, and extreme gradient boosting. Model performance was assessed using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Model calibration was assessed using a calibration curve and a Brier score. The extreme gradient boosting model performed best, with an AUC of 0.7947, although other models showed comparable results. Including community-level SDOH features did not significantly improve model performance overall or within subpopulations. This suggests patient-level variables are the primary drivers of mortality prediction in ESKD. Furthermore, SMOTE did not enhance model performance in subpopulations.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101779"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148555001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence in breast cancer diagnosis: A systematic review of imaging modalities, methods, and datasets","authors":"Shahnaj Parvin,&nbsp;Liton Jude Rozario,&nbsp;Md. Ezharul Islam","doi":"10.1016/j.imu.2026.101786","DOIUrl":"10.1016/j.imu.2026.101786","url":null,"abstract":"<div><div>Artificial Intelligence (AI) has emerged as a transformational tool for breast cancer diagnosis, facilitating swift and precise interpretation of medical imaging while mitigating human error. This systematic review explores AI-driven approaches for breast cancer diagnosis using multiple imaging modalities, including mammography, ultrasound, magnetic resonance imaging, and thermography. In addition to summarizing imaging techniques, this study outlines breast cancer progression patterns and stage classification based on standardized tumor size criteria, providing foundational insights for early-stage researchers. Following PRISMA guidelines, 60 peer-reviewed studies published between 2015 and 2026 were analyzed to evaluate data modalities, algorithmic frameworks, and validation strategies used in AI-based detection systems. This review highlights advances in deep learning models, such as convolutional neural networks, transformer-based architectures, hybrid frameworks, and multimodal fusion techniques. Publicly available datasets and evaluation metrics were also examined. Furthermore, this study proposes an AI-driven multimodal healthcare analytics framework for accurate breast cancer diagnosis. By synthesizing recent advances and practical considerations, this study offers clinicians and researchers a valuable resource on current AI-driven methods for breast cancer diagnosis.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"64 ","pages":"Article 101786"},"PeriodicalIF":0.0,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148641021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}