Artificial-intelligence-based end-to-end prediction of cancer immunotherapy response
Background: Cancer immunotherapy with immune checkpoint inhibitors (ICIs) is widely used in multiple cancer types, with proven benefits. However, response is not guaranteed, difficult to predict, and serious toxicity may occur. Predictive biomarkers for ICIs response exist, but only few of them are clinically used because they require tissue samples, are costly and increase turnaround time. Thus, there is an urgent clinical need to predict response to ICIs at patient’s level. Aims: TANGERINE partners have developed artificial intelligence (AI)-based histology image analysis and computed tomography (CT)-based radiomics for predicting immune features related to ICIs response. We propose to a) expand and combine them to develop and validate an end-to-end open AI tool to predict response and toxicity to ICIs; and b) identify cellular structures and image patterns associated with ICIs response that explain model predictions. Methods: Digital images of tumour histopathology slides and CT scans will be retrieved, linked to clinical outcomes data and anonymized for analysis. An initial retrospective (2017-21) data retrieval from 1800 patients at 6 centres will continue with a prospective recruitment of 600 more to validate models. Patients that received ICls as first line for any tumour will be included and response recorded according to iRECIST. Radiomics and deep convolutional neural networks will be used. Model explainability will use spatial transcriptomics data on a subset of 30 patients. At analysis, homogenous subgroups will be considered, as gender and ethnicity. Expected results and potential impact: TANGERINE will provide a public-available, non-invasive, low-cost tool based on routinely available images and clinical data to accurately predict ICI response and toxicity. The explanatory module might identify new patients on which ICI may be beneficial. The transnational collaboration will provide patients with enough variability to build generalizable models.
Subscribe to our Newsletter You can unsubscribe at any moment by clicking on the subscription preferences link at the bottom of a received newsletter.
This project has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No. 964264.