Models to segment whole slide images into areas of interest, such as carcinoma, stroma, necrosis, and more.
We train these models on our own data, in a supervised approach with expert annotations from trained pathologists.
Models to classify individual cells in H&E, IHC or multiplex IF. We train our models with a proprietary semi-automated approach with millions of labels, leveraging same-section stainings and expert annotations from trained pathologists.
Models to predict molecular or clinical parameters (e.g. outcome, relapse, IDFS), from image data or multi-modal data. We use our patented “Explainable AI” (layer-wise relevance propagation) to reverse-engineer these black-box models and derive features that were learnt by the models, for instance to uncover potential novel biomarkers. We supplement and validate this approach with tissue segmentation and cell classification models.
All of our AI models run in our own proprietary platform, but can also be deployed into third party platforms and in-house platforms of clients. Our platform is available as SaaS solution to our clients and supports a wide range of scanners and image modalities, including H&E, IHC and multiplex IF.