A machine learning model successfully separated brain changes specific to mild cognitive impairment from normal aging, achieving superior prediction of dementia risk. Original paper: Decoupling MCI-specific signatures from shared neurobiological substrates of cognitive aging via deep learning. — NPJ digital medicine. 10.1038/s41746-026-02597-3 📄 Read the article

A 3D foundation model trained on self-supervised learning demonstrates powerful generalization for detecting multiple neurological diseases in head CT, achieving 16-20% accuracy improvement over conventional approaches. Original paper: 3D foundation model for generalizable disease detection in head computed tomography. — Nature biomedical engineering. 10.1038/s41551-026-01668-w 📄 Read the article

An interpretable machine learning model using multi-parameter ultrasound achieves 84% accuracy in discriminating early- versus late-stage ALS, offering an objective, non-invasive clinical tool. Original paper: Predicting amyotrophic lateral sclerosis stage based on multi-parameter ultrasound: development and validation of an interpretable machine learning model. — BMC medicine. 10.1186/s12916-026-04871-3 📄 Read the article

A machine learning model called OsteoSC-M3 can identify women at high risk of osteoporosis approximately 29 months before standard diagnostic confirmation, enabling preventive intervention. Original paper: A multiparameter model (OsteoSC-M3) for early detection and risk stratification of osteoporosis in women: a multicentre cohort study in China. — EClinicalMedicine. 10.1016/j.eclinm.2026.103902 📄 Read the article

A direct comparison of study designs for deep learning models reveals that cohort-based approaches better preserve real-world risk estimates, while case-control designs—despite higher apparent accuracy—systematically overestimate patient risk. Original paper: Cohort vs case-control design for transformer-based prediction of asthma exacerbations in mild asthma. — NPJ digital medicine. 10.1038/s41746-026-02624-3 📄 Read the article

Researchers have identified azole compounds, especially talarozole, as promising drug candidates for Leigh syndrome by combining deep learning computational screening with human brain organoid models derived from patient cells. Original paper: Accelerating Leigh syndrome drug discovery through deep learning screening in brain organoids. — Nature Communications. 10.1038/s41467-026-71391-2 📄 Read the article

A novel two-stage deep learning model significantly improves histological diagnosis of early pregnancy loss, achieving 84.3% accuracy while reducing diagnostic time and substantially aiding pathologist decision-making. Original paper: Patch-to-slide fusion deep learning model for histological diagnosis of early pregnancy loss including hydatidiform mole. — NPJ digital medicine. 10.1038/s41746-026-02628-z 📄 Read the article

A specialized large language model called Xuanwu-NeuroAid outperformed emergency physicians in diagnosing neurological emergencies and significantly improved physician accuracy when used as a decision-support tool. Original paper: Development and prospective shadow evaluation of a domain-specific large language model for emergency neurological diagnosis. — NPJ digital medicine. 10.1038/s41746-026-02644-z 📄 Read the article

Artificial intelligence models developed from gene expression data can accurately predict papillary thyroid carcinoma occurrence and cervical lymph node metastasis risk, potentially enabling more personalized patient care. Original paper: Predictive models for the occurrence and lymph node metastasis of papillary thyroid carcinoma with regional risk heterogeneity. — NPJ digital medicine. 10.1038/s41746-26-00123 📄 Read the article

Your mouth’s bacterial community may hold the key to predicting biological age and disease risk. Researchers have shown that oral microbiome signatures independently forecast mortality and frailty, offering a practical clinical biomarker. Original paper: Oral microbiome signatures predict biological age and host health. — Nature communications. 10.1038/s41467-026-72096-2 📄 Read the article