DiaDeep shared a post on LinkedIn:
“New publication in NPJ Digital Medicine, part of the Nature Portfolio.
In ovarian cancer, HRD status is a key biomarker for PARP inhibitor treatment selection. Yet response heterogeneity remains substantial, even among patients with the same HRD classification.
This is where routine pathology images may add clinically relevant information.
Using data from the phase III PAOLA-1 randomized trial, DiaDeep and its clinical partners developed a treatment-aware deep learning model trained on diagnostic H&E whole-slide images from treatment-naive primary tumors.
The model estimates each patient’s individual benefit from maintenance olaparib plus bevacizumab versus bevacizumab alone.
Key findings from the article include a strong treatment–biomarker interaction, independent association with outcome after adjustment for clinical variables and HRD status, and further stratification of predicted benefit within both HRD-positive and HRD-negative tumors.
At DiaDeep, this is what we are building: AI-derived biomarkers from routine pathology images to help refine oncology treatment decisions and complement existing molecular biomarkers.
Our sincere thanks to the clinical investigators, pathologists and data scientists who made this work possible, and above all, to the patients of PAOLA-1.”
Title: Treatment-aware deep learning enables counterfactual prediction of individual benefit from PARP inhibitors in ovarian cancer
Authors: Jean-Sébastien Frenel, Pierre-Etienne Heudel, Emmanuelle Guinaudeau, Céline Bossard, Carmela Pisano, Giuseppe Caruso, Joseph Rynkiewicz, Sanae Salhi, Yahia Salhi, Jérôme Chetritt, Eric Pujade-Lauraine, Isabelle Ray-Coquard
Jean-Sébastien Frenel: Can AI Predict Which Ovarian Cancer Patients Benefit From PARP Inhibitors?
Other articles about ovarian cancer on OncoDaily.