The AI

In the next decade of personalized medicine in cancer and pathology, AI will play a pivotal role in translating “super-human” complex patterns into actionable insights. To achieve this, the recipe is clear: big data, big AI expertise all combined with in depth medical expertise. But what does “big” even mean? For instance, is it realistic to expect a pathologist to identify millions of cells in just a single gigantic H&E slide to train the AI algorithm, let alone repeat that for thousands of slides? This unrealistic level of “big data”, is the guidance today’s machine learning algorithms require.Algorithms requiring obscene amounts of data should not come as a surprise. Expert pathologists take years and even decades of dedicated training to perform their delicate duties after all. For an expert pathologist all this training is absolutely necessary to understand a biopsy in detail, to deduce, to make expert judgements from limited data; in short, to extrapolate. By contrast, today’s machine learning algorithms are trained to memorize their big data. In other words, they learn to interpolate and cannot easily extrapolate. Under the leadership of expert pathologists and medical researchers at the world-class Netherlands Cancer Institute, and of expert professors of AI at the University of Amsterdam, Ellogon.AI has the answer. Our proprietary and groundbreaking Exclusive Deep Learning Technology does not just blindly copy the pathologist but extrapolates.

By exploiting sparse-shot and self-supervised neural dynamics during the extremely high-dimensional optimization, Ellogon’s Exclusive Deep Learning identifies millions of cells and “super-human” patterns in raw H&E and PD-L1 stained histopathology images. Ellogon’s Exclusive Deep Learning can extrapolate because it explicitly excludes all biopsy information that is unrealistic for the pathologist to interpret and detrimental for the machine to train with. Instead, with a self-learning strategy Exclusive Deep Learning deduces and incorporates knowledge automatically and indirectly. Ellogon’s Exclusive Deep Learning technology procures algorithms of high accuracy and robustness, quantifying relevant immunogram biomarkers including tumor infiltrating lymphocytes (TIL’s), PD-L1 inhibition checkpoints, and Tumor Mutation Burden (TMB) associated biomarkers. With high-quality immunogram quantification, Ellogon informs the pathologist of the specific biomarkers per individual patient, thus enabling personalized immunotherapy plans. Ellogon’s technology is validated thoroughly in multi-center medical data and cohorts worldwide to meet CE-certified standards of excellence and quality according to current regulation standards.