PathAI, a Boston-based AI technology company, has found that AI can help specialists quantify PD-L1 expression on both tumor cells and immune cells.
Data was collected from more than 1,000 patients who participated in three Phase 3 clinical trials aimed at evaluating the effectiveness of Opdivo, a cancer medication manufactured by Bristol-Myers Squibb, in patients with non-small cell lung cancer. The researchers’ algorithm was trained using more than 250,000 annotations provided by pathologists using PathAI’s own research platform. The performance of AI-powered quantification was compared to that of manual quantification by a team of five pathologists, and the two methods “found similar associations” between PD-L1 expression and progression-free survival. The full results will be presented at the Society for Immunotherapy of Cancer (SITC) 2019 annual meeting November 9, 2019, in National Harbor, Md.
“Together with partners like PathAI, Bristol-Myers Squibb’s translational scientists are tapping into the most advanced imaging, digital pathology and AI technologies available in order to accelerate research—at scale—that can help predict patient outcomes more quickly and accurately,” Saurabh Saha, MD, PhD, Bristol-Myers Squibb’s senior vice president and global head of translational medicine, said in a prepared statement. “The combination of our unique expertise and sophisticated capabilities enables us to explore biological insights in real-time and translate them into meaningful learnings that fuel our R&D organization.”
In a separate study, also set to be presented Nov. 9 at SITC 2019, the two companies explored data from nearly 300 patients with urothelial cancer and found that PathAI’s research platform “performed similar to or better than pathologist-based scoring in all cell types tested, and especially for immune cells, where manual correlation is known to be low.”
“In both studies, we put our process up against a very rigorous and high-quality benchmark,” Andy Beck, co-founder and CEO of PathAI and an author of both studies, said in the prepared statement. “We’re pleased to see that the PathAI approach performed well, demonstrating performance at the level of board-certified pathologists exhaustively labeling cells by hand across the board.”