Medical Image Classification Using Deep Learning

Abstract

Image classification is to assign one or more labels to an image, which is one of the most fundamental tasks in computer vision and pattern recognition. In traditional image classification, low-level or mid-level features are extracted to represent the image and a trainable classifier is then used for label assignments. In recent years, the high-level feature representation of deep convolutional neural networks has proven to be superior to hand-crafted low-level and mid-level features. In the deep convolutional neural network, both feature extraction and classification networks are combined together and trained end-to-end. Deep learning techniques have also been applied to medical image classification and computer-aided diagnosis. In this chapter, we first introduce fundamentals of deep convolutional neural networks for image classification and then introduce an application of deep learning to classification of focal liver lesions on multi-phase CT images. The main challenge in deep-learning-based medical image classification is the lack of annotated training samples. We demonstrate that fine-tuning can significantly improve the accuracy of liver lesion classification, especially for small training samples.