What happened
Researchers built a system that keeps patient data private by processing it locally, then only sends small summaries to a central server — and it's smarter about which computations to actually run, skipping expensive image analysis when the easier data alone gives a clear answer. This means medical AI can run on battery-powered edge devices without draining power or requiring data to leave the clinic.
Why it matters
Medical AI deployed at the edge (in hospitals, clinics, devices) currently either wastes power on unnecessary computation or sends raw patient data over networks to central servers. This shows a path where systems can be both private and efficient by making routing decisions based on actual clinical need — turning medical inference from 'process everything' into 'process only what matters.'