Residual block Convolutional Auto Encoder in Content-Based Medical Image Retrieval

Approximately 12 percent of men will suffer from prostate cancer during their lifetime. While some types of this cancer grow slowly and need minimal treatment, others are aggressive and spread quickly. In order to be able to provide a fast and accurate diagnosis, computer vision techniques are widely used to help pathologists in the diagnosis and prognosis tasks. In this work, a new Content-Based Medical Image Retrieval (CBMIR) method is presented. The method, which is named ResCAE, presents a modified Convolutional Auto-Encoder (CAE) with a residual block and a skip layer to extract the relevant features of prostate cancer in Whole Slide Images (WSIs) in SICAPv2 data set. To the best of the author’s knowledge, this is the first study of using a Res-CAE to perform a search engine in prostate cancer images. After extracting features, Euclidean and Cosine have been applied to calculate the similarity of the query image and the gallery. The experimental results show that the proposed method achieves 0.85, 0.78 of accuracy by computing 7-top and 5-top accuracy, respectively.