چكيده به لاتين
Abstract
The analysis of medical images play a dominante role in the detection of many diseases, the automatic analysis of these images can help the doctor to detection, and also reduce workload. Deep learning is one of the most popular methods for analyzing medical images. Among the various approaches in this field, the Convolutional Neural Network (CNN) has become popular, due to the fact that it has a high ability in automatically extracting high-level features from images. In this thesis, we proposed systems for the classification of liver cancer Computed Tomography (CT) scans which used in their design. The CNN, due to the low contrast of the CT scans, was unable to extract the proper features for classification. Hence, we improved the contrast of CT scans by image processing techniques to extract better feature. Finaly, these features have been classified by Multi-Layer Perceptron (MLP) and Suport Vector Machine (SVM). We trained and tested proposed systems with 560 CT scans of liver cancer which were a subset of TCGA-LIHC. The best proposed system, using the stretch and Equalization histogram technique with removing the margins of the scan, has 88.125% rectitude in Classification of CT scans reached liver cancer.
Keywords: deep learning, neural network, image processing, computed tomography scan, liver cancer
