GE HealthCare-RadNet AI Mammography Expansion and AI Workflow Advances
Key Questions
What AI advancements are GE HealthCare and RadNet implementing for mammography?
GE, RadNet, and DeepHealth are scaling AI tools to improve detection in dense and diverse breast tissue, achieving a 21% improvement. New models include an MRS texture approach with a hazard ratio of ~2.3 that generalizes across populations and a multimodal deep learning system with 0.973 AUC to reduce unnecessary biopsies.
What results did the large Swedish AI mammography study show?
The study of over 100,000 women confirmed that AI-assisted reading detects more cancers without increasing recall rates. This supports broader adoption of AI in screening workflows.
How is federated multi-modal deep learning being used in breast cancer diagnosis?
A privacy-preserving federated model combining ultrasound, MRI, and mammography data achieved 93% fusion accuracy. It enables collaborative training across institutions without sharing patient data.
What funding did Subtle Medical secure and how will it be used?
Subtle Medical raised $33M to advance AI-driven MRI throughput, adding 4-5 extra scan slots per scanner daily. RadNet is a key customer supporting these efficiency gains.
How is UCSF applying AI to reduce wait times for high-risk breast screening?
UCSF developed an AI triage tool that cuts wait times for high-risk women from weeks to hours at a safety-net hospital. The effort focuses on equity while addressing bias risks in developing systems.
GE/RadNet/DeepHealth scaling AI for 21% dense/diverse detection; new MRS texture model generalizable across populations (HR~2.3) and multimodal DL (0.973 AUC) reduce biopsies/false positives. Hologic/Röko advances noted. Large Swedish AI study (100k+) confirms AI-assisted reading detects more cancers without higher recall. New federated multi-modal DL (US, MRI, mammo) achieves 93% fusion accuracy with privacy-preserving approach. Subtle Medical raises $33M for AI-driven MRI throughput (4-5 extra slots/scanner/day) with RadNet as customer. UCSF develops AI triage tool cutting wait times for high-risk women from weeks to hours at safety-net hospital. Status developing amid bias/equity risks.