Artificial Intelligence Analysis of Histopathology Slides to Develop Biomarkers of Response to Immunotherapy in Kidney Cancer
Abstract
Cancer immunotherapy is one of the most promising areas of treatment for cancer currently. Immunotherapy is a type of treatment that uses the patient s own immune system to fight their cancer. Breakthrough treatments have been discovered in the last decade leading patients into almost complete remission of their cancer. However, not all patients respond to immunotherapy. Therefore, our ability to predict which patient will respond to immunotherapy before they receive the drug will enable treatment of patients who are more likely to respond. The focus area of this project is kidney cancer, where ~50% of metastatic patients receive immunotherapy. However, there are no biomarkers for patient selection in kidney cancer for immunotherapy. Specifically, in kidney cancer, there is an urgent need to identify biomarkers for selecting patients most likely to respond to immunotherapy. Prediction of response to drugs is an area of active research. In this project, we plan to utilize histology slides. Histology slides are slides of a slice of the tumor biopsy of the patient. Typically, these slides are utilized primarily for cancer diagnosis, namely, presence or absence of tumor, type of cancer, and the stage of the tumor. However, they have been shown to also contain a wealth of additional information such as which cells are present in and around the tumor and other biological features associated with the tumor. One approach to understand and analyze these images further is by the use of artificial intelligence (AI). While recent advances in AI can accurately predict kidney cancer subtypes, the use of AI to predict response to immunotherapy has not been explored. In this project, we propose that an investigation of the patient s tumor biopsy using image analysis will allow us to better predict which patients will respond to immunotherapy by learning and characterizing the features of the tumors in kidney cancer, namely, the spatial arrangement of the various cell types, the presence of immune related cells, the presence of dead cells, and other features in the tumor. Our goal is to develop biomarkers of immunotherapy response using AI models from histology images of the tumor biopsies. We will first use AI to identify immune cell types on histology images, and then utilize the 2-D spatial patterns of these immune cells as features that could potentially predict response to immunotherapy. For the first part of this work, we will use multiplex immunofluorescence (mIF) to classify the immune cell types seen in the histology slide. mIF technologies, which allow the simultaneous detection of multiple (as many as 38) markers on a tissue section, have been increasingly used in research and clinical work in response to increased demand for better techniques to characterize tumor biopsies, allowing for comprehensive studies of cell composition. This knowledge can then be used for diagnosis as well as for biomarker discovery. In this project, in the first part, for each cell type, we will develop a separate AI model using mIF as training data to predict a probability map of the presence of that cell on the original histology image. The result of the model will be a probability map prediction for each of the immune cell types. These probability maps will serve as input sets for predicting response to immunotherapy. In the second part of the project, we extract latent features from the AI model and will utilize genomic and clinical features of the tumor to interpret these predictive features. Our innovation is in utilizing histopathology for therapeutic biomarkers. Our proposal has the potential to change the current paradigm of utilizing histology mainly for diagnosis, to also guide therapeutic decisions for kidney cancer patients. Our study also paves the way for interpretable AI-models and showcases the value for a new data type: histomics.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 28, 2022
- Source ID
- W81XWH2210367
Entities
People
- Anupama Reddy
Organizations
- United States Army