Open-source AI for rapid disease detection

The deep learning model could help pathologists spot and monitor cancers

Spotted: Pathologists play a central role in the diagnosis and prognosis of life-threatening diseases like cancer, using their expertise to closely examine organs, tissues, blood, and bodily fluids to identify notable abnormalities and monitor changes as a patient undergoes treatment. It’s a difficult and often highly specialised role, but French startup Bioptimus believes AI can lend a hand.

The company has developed a large open-source foundation AI model called H-optimus-0, which could help pathologists make quicker and more accurate diagnoses. As doctors around the world face a growing number of increasingly complex cases, the model could help to ensure that diseases and disease progression are caught early on, enabling quicker interventions and better patient outcomes.

H-optimus-0 has been trained on a proprietary dataset that includes hundreds of millions of images from over 500,000 histopathology slides (tissues or cells under a microscope). These images come from over 200,000 patients across 4,000 different clinical practices, and because the deep learning model has been trained on such a comprehensive and diverse set of data points, it means it can respond effectively to a range of scenarios, including identifying specific tissue characteristics and detecting relevant biomarkers or the spread of cancerous cells.

The cutting-edge model was announced in July, just five months after Bioptimus emerged from stealth following a $35 million seed funding round. According to company Co-Founder and CEO Jean-Philippe Vert, “H-optimus-0 is just the beginning (…) Future models will not only be trained on an even larger number of pathology images from Europe, Asia, and Africa but will also incorporate other modalities, such as genomics and proteomics.”

Written By: Matilda Cox