Medical imaging articles from across Nature Portfolio

AI models for medical diagnostics are vulnerable to membership inference attacks.

Pu et al. use machine learning to quantify peripheral hard drusen across nine retinal quadrants in 1,562 normal eyes using indocyanine green angiography. Nasal quadrants show earlier and greater drusen accumulation, and central superior drusen count independently predicts progression to neovascular age-related macular degeneration.

Algorithmic advances enable ultrasound localization microscopy with clinical ultrasound systems, facilitating bedside assessment of microvasculature in people with traumatic brain injury and revealing that vascular remodelling is predictive of long-term outcomes.

Coronary artery calcium visible on routine, low-dose, non-contrast, chest computed tomography (CT) scans provides clinically relevant information about cardiovascular risk, without additional radiation exposure or cost. Mandatory systematic reporting of coronary artery calcium findings, supported by artificial intelligence and standardized workflows, could transform and elevate an often-overlooked, incidental finding into a practical tool in preventive cardiovascular care.

Symptomatic knee osteoarthritis at an early structural stage could provide a window of opportunity for intervention. Symptomatic knees without definite radiographic osteoarthritis can already show structural joint changes detectable on MRI; however, the greater sensitivity of MRI must be balanced against the risks of overdiagnosis and overtreatment.

Using large-scale retinal images from community and tertiary hospitals, we developed Reti-Pioneer, a quality-aware, multi-task framework for multi-disease detection. Diverse external validation confirmed its generalizability. Furthermore, a prospective silent trial and clinical pilot study demonstrated its time efficiency, real-world feasibility, and potential for integration into clinical workflows.

This Comment announces the Precision Mental Health Commission and its initiative to redefine mental disorders through brain circuit function analysis, promoting stratified, circuit-informed care that enhances treatment accuracy and efficiency. The approach mirrors advances in cardiovascular and cancer research, moving beyond trial and error toward personalized care.