Periodic Reporting for period 1 - AI-PATH (Artificial intelligence expert to boost growth of a high-potential digital diagnostics start-up)
Reporting period: 2020-10-01 to 2022-01-31
Cancer is the second leading cause of death globally. Immunotherapy is a Nobel Award winning therapy for cancer which enables human's own immune system to attack and kill cancer cells. Immunotherapy shows great results but has a low treatment efficacy. This means that the survival rate of cancer patients that respond to immunotherapy is very high compared to other treatments but only a small fraction of patients actually respond to it. At the same time immunotherapy is an expensive treatment increasing the financial burden of healthcare systems around the world and despite the low toxicities still has significant side effects. Therefore the question that naturally arises is how can we identify the patients that have the highest chances to respond to immunotherapy. To this date the selection of patients is a tedious, time-consuming, erroneous manual task that pathologies and oncologists perform manually. In particular, pathologists look into cancer cell biopsies to quantify the probability of the immune system attacking the cancer cells. The objective of the AI-PATH project is to empower pathologists and oncologists by letting intelligent algorithms look into digital biopsies and identify some of the important signals that point towards the use of immunotherapy. One of such signals is whether there are enough lymphocytes close by the tumor cells.
The work performed within the AI-PATH project is to develop tools that can quantify the number of lymphocyte cells that are located around the tumor cells in relation to the stroma around the tumor cells. It is well-known that if that ratio (named sTIL) is higher than a threshold, i.e. if there are enough lymphocytes in the vicinity of the tumor the chances of killing the tumor cell are higher. The problem in mathematical terms is a problem of detecting very objects in an image, similar to detecting cats or dogs in a picture. However these objects (lymphocytes and tumor cells) are very small and they all look alike to a non-expert's eye. Therefore we need to train intelligent algorithms to recognise the different types of cells. We use existing state-of-the-art algorithms and train them on publicly available datasets yielding a measure of accuracy (F1) of 65%. This was not good enough. What is the problem? An expert pathologists needs to tell the algorithm which cells are lymphocytes, which cells are tumor cells and which cells are none of the two. However a biopsy contains millions of cells and the pathologist does not have the time to indicate what each of these cells are. Therefore the algorithm has a lot of missing (or to some extend misleading) knowledge. In this project we worked on informing the algorithms about these misleading training data, which yield an accuracy measure of 80%, very close to the performance of a pathologist, but much faster and in a bigger scale. That is, while the pathologists has the ability to zoom in with his microscope to very few regions of the biopsy to find cells, our algorithm finds these cells everywhere in the biopsy making the final scoring of the sTILs more reliable. At the same time during this project we developed visualisation tools to qualitatively assess the output of the algorithm.
In conclusion we built the models, algorithms, tools and software to quantify this important biomarker faster and more accurately than a pathologist and present our finding to a pathologist for her/him draw conclusions upon it for whether immunotherapy has good chances of success. There is no state-of-the-art on this problem. We were the first to solve and set the state-of-the-art. Our findings where published but also verified by collaborators on proprietary datasets (to which we do not have access) and made robust and generalizable. The results of the AI-PATH project helped Ellogon AI secure loans and funds that allowed the growth of the company from 0 FTE (prior to AI-PATH) to 7 FTE (5 full time employees, 2 part-time free-lancers, and 3 part-time advisors). For our project to have a practical implication in the field of medicine we need to clinically validate the results in a retrospective experiment, which we have set for May 2022. If the results are positive then we can offer our algorithm for clinical use.