Paper Title
Prostate Segmentation in T2-Weighted, Apparent Diffusion Coefficient and Diffusion-Weighted Imaging MRI Scans

Abstract
According to World Cancer Research Fund International prostate cancer is the second most common following lung cancer and the fifth most common cause of cancer death amongst men. Prostate cancer is also the fourth most frequent tumor following lung, breast and colorectal cancers between both genders worldwide. Early prostate cancer usually has no symptoms, leading to the need of screening to detect it in early stage. Biopsy, if performed on the right location, is the only way that has the ability unambiguously detect prostate cancer so far. Despite that, biopsy is able to detect only 70-80% of clinical cancer cases. Nowadays in order to determine the location to perform biopsy on multi parametric magnetic resonance imaging (mpMRI) is used. Most often manual localization takes a lot of time and is inaccurate even when using mpMRI. Software that is able to perform automated prostate and abnormal prostate areas localization in standardized manner is needed to tackle this issue. Standardized methods while examining patients enable effective collaboration between radiologists and urologists. Despite this, algorithms being closed source or hard to implement without the help of the original author lead to troublesome comparison of effectiveness of different algorithms. The problem is further aggravated by non-standardized MRI signal intensity, acquisition protocol, coil profile, field strength and scanner type.Prostate MR Image Segmentation (PROMISE12) challenge tries to tackle this issue by providing common dataset of images and same technique for calculating results for comparing effectiveness of different algorithms. Authors chose the open source algorithm that showed best results in PROMISE12 challenge (“no new” U-Net – nnU-Net) and investigates its performance under different circumstances, e.g. by introducing sets of images. Authors discuss the algorithm’s segmentation results changes when using couple of slices rather than full image series for training as well as including mpMRI apparent diffusion coefficient and diffusion-weighted imaging modality images in addition to T2-weighted used in original PROMISE12 challenge. Keywords - Prostate segmentation, MRI, T2-weighted scan, apparent diffusion coefficient, diffusion-weighted imaging.