Scientists at Harvard Medical School have designed an AI model similar to ChatGPT to assist in cancer diagnosis and treatment decisions. The AI model, named Clinical Histopathology Imaging Evaluation Foundation (CHIEF), detects cancer cells by analyzing digital slides of tumor tissues. Additionally, it predicts survival rate and identifies patients who may benefit from experimental treatments.
CHIEF, similar to CHATGPT, is versatile by performing various tasks across multiple cancer types, including lung, breast, prostate, and more. It was trained on 15 million unlabeled images and 60,000 whole-slide images of tissues. Moreover, it was tested on more than 19,000 images from 32 global datasets. Thus, unlike current AI models that are limited to certain populations, this new technology has the ability to predict patient outcomes and validate them across several international patient groups. Furthermore, its remarkable accuracy in image assessments surpasses that of many current AI systems.
It was tested on 19 different cancer types and achieved a nearly 94% accuracy in cancer detection. A prospective clinical study to validate the CHIEF model in real-world clinical settings is currently underway. Its adoption will position it as a powerful tool for oncologists worldwide, further advancing the notion of global inclusivity in AI.