Discovering New Cancer Imaging Biomarkers with Foundation Models: A Game-Changer in Radiology


Researchers at Mass General Brigham have employed foundation models, the technology behind AI tools like ChatGPT, to unearth new cancer imaging biomarkers. These biomarkers have the potential to revolutionize the way radiological patterns are identified from medical images, significantly enhancing early cancer detection and treatment.

The team developed their foundation model using a vast dataset of 11,467 radiologic scans with abnormalities. By analyzing these images, the model successfully identified patterns that could predict anatomical site, malignancy, and prognosis in three distinct use cases across four patient cohorts.

Compared to traditional methods, this innovative approach demonstrated impressive performance even when applied to niche tasks with limited available data. The team’s findings have been published in the journal Nature Machine Intelligence.

Given the growing specificity of image biomarker research, the researchers believe their work will pave the way for more precise and efficient investigations. However, a critical concern remains the reliability and explainability of AI methods, as their answers must be understandable to human experts.

1.  Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it