A multi-level deep learning algorithm to estimate tumor content and cellularity of prostate cancer

Ahmed Serag, Peter Hamilton, Perry Maxwell, Paul O'Reilly

Abstract:

Prostate cancer is the most common cause of cancer death in men in the UK, responsible for more than 11,000 UK deaths each year. Diagnosis and grading of prostate cancer is an increasingly complex process and requires a more detailed
analysis from pathologists to complete it with accuracy. Such process is not only time-consuming, but also prone to intra- and inter-observer variabilities. We designed a multi-level deep learning algorithm for the estimation of tumor content
and cellularity of prostate cancer using fully convolutional neural networks. The approach provides accurate and consistent results, and is applicable across the whole range of computational pathology problems.

Keywords: artificial intelligence, deep learning, cancer detection, prostate, computational pathology, digital pathology

Author Affiliation: Philips Digital Pathology Solutions, Philips Digital Pathology Solutions, Queen’s University Belfast, Philips Digital Pathology Solutions