Deep learning for prostate cancer
Title: Deep learning for prostate cancer
DNr: Berzelius-2024-166
Project Type: LiU Berzelius
Principal Investigator: Eduard Chelebian Kocharyan <>
Affiliation: Uppsala universitet
Duration: 2024-05-01 – 2024-11-01
Classification: 30203


The goal of the project is to predict whether a patient will develop prostate cancer in the future based on prostate needle biopsies. For that, we have a database of biopsies that were deemed benign at time=0. However, some of them developed cancer in a span of 2 years (label=1), while others did not even after 10 years (label=0). The deep learning techniques provided below could help prognostication and see if the physician missed the malignancy or if there was something else that indicated that a particular patient would develop cancer.