Improving Prostate Cancer Diagnostic Pathway with Transformers and a Generative Self-Supervised Approach
Abstract
Prostate cancer is the second most occurring cancer and the sixth leading cause of canceramong men worldwide. The number of cases are expected to increase due to populationgrowth and increase in expected lifetime. The screening test for prostate cancer is notwithout risk and there is chance of over-diagnosis with the traditional screening. Themagnetic resonance imaging (MRI) examination is an essential tool and a comfortablemethod for diagnosing cancer. If some of the screening methods can be replaced byaccurate MRI examinations it will be more comfortable for the patients.
This thesis will explore how masked autoencoders with a generative self-supervisedapproach can improve diagnosis of prostate cancer from MRI images. Computer vision isused in many ways in the field of medical imaging. Hopefully computer vision based ondeep learning and transformers can improve the field of medical imaging. However, thestate of the art deep learning models require pre-training on huge amounts of unlabeleddata and fine-tuning on large amounts of data. This thesis investigates if a masked visiontransformer can learn to predict prostate cancer lesions on a limited amount of data. Thefinal results suggest that a simple model of small size is not sufficient to for predictionprostate cancer.