A Deep Learning Approach for the Automatic Detection and Segmentation in Autosomal Dominant Polycystic Kidney Disease Based on Magnetic Resonance Images

The automatic segmentation of kidneys in medical images is not a trivial task when the subjects undergoing the medical examination are affected by Autosomal Dominant Polycystic Kidney Disease. In this work, two different approaches based on Deep Learning using Convolutional Neural Networks (CNNs) for the semantic segmentation of images containing polycystic kidneys are described, leading to a fully-automated classification of each pixel of the images without the need to extract hand-crafted features. In details, the first approach performs the automatic segmentation of the kidney considering the whole image as input, without any pre-processing. Conversely, the second approach is based on a two-steps classification procedure constituted by a CNN for the automatic detection of Regions of Interest (ROIs), according to the R-CNN approach, and a subsequent convolutional classifier performing the semantic segmentation on the ROIs previously extracted. Results for both the approaches are reported considering different metrics evaluated on the test set. Even though the R-CNN shows an overall high number of false positives, the subsequent semantic segmentation on the extracted ROIs allows achieving good performance in terms of mean accuracy. Moreover, the results show that both the investigated approaches seem to be reliable for the automatic segmentation of polycystic kidneys, as in both the cases an accuracy of about 85% was reached.