Harvard Develops Versatile AI for Cancer Diagnosis and Prognosis

The new AI model, CHIEF, demonstrates high accuracy across multiple cancer types and tasks, outperforming existing systems.

CHIEF was trained on 15 million images and tested on 19,400 images from 32 global datasets.
It can detect cancer cells, predict tumor origins, and forecast patient outcomes with up to 94% accuracy.
The AI model identifies genetic mutations and DNA patterns, providing a cost-effective alternative to DNA sequencing.
CHIEF's adaptability allows it to work with biopsy and surgical samples, regardless of digitization technique.
Future enhancements aim to include rare diseases, pre-cancerous cells, and predict novel treatment effects.