AI-based digitised pathology to identify subtypes in breast cancers

Lead Supervisor
Dr Anita Grigoriadis
Senior Lecturer
School of Cancer and Pharmaceutical Sciences, King’s College London
anita.grigoriadis@kcl.ac.uk

Co-supervisor
Professor Sarah Pinder (King’s College London)

Industry Partner
Pandu Raharja-Liu (Panakeia Technologies)

Project Details

Unprecedented advances in machine learning methods have enabled the synergy of artificial intelligence and digital pathology, which offers image-based diagnosis possibilities that were once limited only to radiology. In this project, we will use digitised whole slide images (WSI) from a very aggressive breast cancer type, namely triple negative breast cancer (TNBC). Patients of TNBC have currently limited treatment options, as this type of cancers is clinically and molecularly very heterogeneous. So far, we have only identified few subgroups of patients for whom targeted therapies are available.

Our objectives are to implement machine learning methods to capture morphological patterns in TNBC based on hundreds of digitised H&E. Moreover, we will expand these models by integrating available clinical data, information of the presence or absence of known breast cancer biomarkers (e.g. Ki67, Keratin 14, Keratin 5, EGFR) and omics data (such as gene expression, copy number data). Our aim is to create models which may reveal subgroups within TNCB patients, who have different risk of developing distant metastasis, treatment response and overall survival.

The student will be trained in digitised image analyses, machine learning methods and outcome analyses.

Datasets

The data available for this thesis will comprise in-house data sets, as well as publicly available TNBC cohorts with available histopathological, clinical, and molecular information, such as the TCGA and METABRIC 

Keywords

machine learning, AI, digital pathology, breast cancer subtypes