10 - Deep neural network in medical image processing

Abstract

Computer-assisted analysis for better interpretation of biomedical images has been a long-standing concern in the field of medical imaging. On the image interpretation front, recent developments in machine learning, especially in the field of deep learning, have made a major leap forward to help with the identification, classification, and quantification of patterns in biomedical images. Specifically, the leveraging of hierarchical structures based entirely on data, instead of handcrafted features mostly based on domain-specific knowledge, lies at the heart of the advances. In this way, deep learning is rapidly proving to be the state-of-the-art platform for achieving superior performance in various biomedical applications. The most important point in efficacious treatment of a disease is an accurate diagnosis, which by and large is not an easy feat to achieve. There are many reasons for an inaccurate diagnosis such as inexperienced doctors, scarcity of trained radiologists, lack of proper equipment, and so on. But technology (specifically machine learning) provides a silver lining, as it has been proven quite successful in aiding and assisting diagnosis by interpreting biomedical images such as X-ray, computerized tomography scans, mammograms, and so on. The unprecedented success of deep learning frameworks in the interpretation of biomedical images is mainly due to the following factors: