Chapter 17 - Natural Language Processing for Large-Scale Medical Image Analysis Using Deep Learning

Abstract

Recent advances in deep learning enable us to analyze a large number of images efficiently; however, collecting such large dataset has been mostly hindered by the rate of manual human efforts. Nonetheless, medical images are usually saved with the accompanying radiology reports, and accommodating the natural language information for image analysis has great potential. For example, data collection can be automated to leverage the large volume of data available in the Picture Archiving and Communication Systems (PACS). Additionally, image annotation can be automated by incorporating the human annotation in the radiology reports. The size of medical dataset usually is much smaller than the natural image dataset which advanced deep learning technology is developed for. We can unleash the full capacity of deep learning for analyzing a large volume of medical images, by automating the data collection and annotation. Moreover, a sustainable system can be developed even when the data are continuously being updated, shared, and integrated. This chapter will review some fundamentals of natural language processing (NLP) and cover various NLP techniques to help automate medical image collection and annotation.