Deep Learning Techniques to Classify and Analyze Medical Imaging Data

Abstract

• In recent years, deep learning techniques particularly Convolutional Neural Networks (CNNs) have been used in various disciplines. CNNs have shown an essential ability to automatically extract large volumes of information from big data. The use of CNNs have significantly proved to be useful especially in classifying natural images. Nonetheless, there have been a major barrier in implementing the CNNs in medical domain due to lack of proper training data. As a result, general imaging benchmarks such as ImageNet have been popularly used in the medical domain even though they are not so perfect as compared to the CNNs. In this paper, a comparative analysis of LeNet, AlexNet and GoogLeNet have been done. Thereafter, the paper has proposed an improved conceptual framework for classifying medicinal anatomy images using CNNs. Based on the proposed design of the framework, the CNNs architecture is expected to outperform the previous three architectures in classifying medical images.