Karen Taylor, PhD

Utilizing artificial intelligence to decipher patterns in breast tumor cells that may indicate how they respond to different treatments.

Impact

Pathologists examine the patterns of cells and the presence of biomarkers to diagnose and subtype breast cancer. Examination of patient samples may also help delineate other important details about the type of cancer, its severity, and what treatments might work. However, current methods rely on human assessment which makes the detection of these details subjective and somewhat limited. Artificial intelligence (AI) algorithms are computer-based techniques that can be trained to detect patterns that provide insights into factors not accessible to the human eye. Studies that seek to use AI technology face a key challenge in that they lack large sample sets necessary to validate findings critical to driving change in clinical management. This slows progress in this field which has the potential to directly impact patient care. In this project, Dr. Taylor and her colleagues will utilize AI to examine a large cohort of patients’ samples in order to identify features of the tumor and its environment that might predict a patient’s outcome and response to treatment.

What’s next

The University of Edinburgh has a database of over 22,000 patients’ samples from multiple breast cancer clinical trials and outcome data for these patients is available from the Oxford EBCTCG (Early Breast Cancer Trialists Co-operative Group). The EBCTCG has played a pivotal role in the validation of multiple treatment approaches to the management of early breast cancer. Dr. Taylor’s team will link these resources to build a large library of digital images that can be analyzed using AI algorithms. They hope to identify new patterns in the tumor cells and the surrounding tissue that inform how well patients do and how well their breast cancer responds to different treatments. This large dataset will be leveraged to validate their findings and the utility of AI technology to aid pathologists in routine reporting as well as potentially provide new insights that will enable doctors to make better individualized treatment choices.