Deep Learning for Large-Scale Analysis of Tumour-Infiltrating Lymphocytes on Whole-Slide Images.

点击翻译按钮获取中文摘要

Background: Cancer is a significant contributor to global morality, with an increasing number of cancer patients each year, our healthcare systems have been under huge pressure. There is an urgent need to improve the accuracy and the efficiency of cancer diagnosis, prognosis and the quality of treatment plans. In recent years, the study of tumour microenvironment (TME) has received lots of attention. One type of cells in the TME are Tumour-Infiltrating Lymphocytes (TILs), which are a type of immune cells that can play a part in killing tumour cells, measuring TILs can help better target treatments, particularly immunotherapy. Aims and Objectives: This project aims to develop fully automated deep learning models that can detect TILs on whole-slide images of varies cancer types and measure the extent of TILs infiltrate. The project also aims to predict treatment outcomes based on TIL scores.Novelty of research methodology: We will build our deep learning models on top of existing state-of-the-art models. We will also provide accessible user interface which will enable the models to be integrated seamlessly into pathology workflow.Research impact: Using AI to detect and measure TILs on a large scale will massively speed up pathology workflow, allowing diagnosis and prognosis to be produced faster. Predicting treatment outcomes based on TILs score will allow the optimal treatment to be selected for each patient, therefore maximising survival rate.Alignment to research council strategies and research areas: This project falls into the EPSRC's research area of "AI for healthcare", it supports the strategy "Securing better health, ageing and wellbeing".