Deep Learning Applications in Medical Imaging: Introduction to Deep Learning-Based Intelligent Systems for Medical Applications

Introduction

Deep learning (DL) and artificial intelligence (AI) is becoming an emerging tool in the healthcare sector for medical diagnostics. This chapter elaborates on general reasons for the popularity of computational techniques such as deep learning, Machine learning (ML) applications in the medical image processing domain. The emergence of exploratory and confirmatory DL based decision making is the need for the most developed and developing countries. This fertile field of research is now the area of interest for multidisciplinary people such as computer science, biomedical, biochemistry, bioinformatics etc... However, the practical applications of DL in clinical diagnosis techniques are relatively small, with many challenges added to it. The initial part of this chapter focuses on reviewing the fundamental concepts of the neural network's algorithms. The ML versus DL algorithm is elaborated in brief. The major players in healthcare industries with AI support are also been discussed. The performance and assessment of different architectures for deep neural networks are discussed using the open-source software. Doing so allows us to understand the reasons for the construction of intelligent architectures like multilayer, feedforward, recurrent algorithms. All of them serve a different purpose and offer a different view on various domains in the medical sector. The tools and technology required for DL, selection, Implementation, optimization, testing are discussed with respect to an application of cancer detection. Thus this chapter giver overall vision of deep learning concepts related to biomedical research.

In today’s vision, observably medical image processing is one of the thirst areas which have been largely intervening domain progressing back again with DL. The impact of the DL algorithms and models in healthcare domain has narrowed down its way for data mining, analysis and has also attracted attention in the area of medical image analysis. At present, DL is been employed to involve the classification problems in medical image segmentation, recognition, and brain function studies. The images mainly include ultrasound, X-rays, computed tomography (CT), nuclear magnetic resonance (MRI), digital blood vessel silhouette (DSA), and positron emission tomography (PET). In the field of medical image analysis, the main research directions are image segmentation, image registration, and information fusion, time-series image analysis, and content-based image retrieval. Hence, it is important to select methods of significance and describe them in high detail which is served in this chapter.

Added on, how to effectively perform medical image analysis is a major problem in the field of disease diagnosis, and further diagnostic methods in pieces of literature are also discussed in this chapter. At this stage, DL could be viewed as a black box requiring knowledge of its internal workings, and hence presents some crucial technical challenges that need further methodological development techniques which are critically reviewed. By looking toward the contribution of DL in today’s era, researchers are thinking that by the next 20 years, DL will serve for full automation for all kind of cyclic works. The deployment of DL models gives great benefits in healthcare.

Figure 1.

Schematic representation of Network models and its applications

As an overall summary, first, the chapter has initially placed the basic requirements of the DL algorithm over machine learning architectures, followed by the understanding of all architectures along with its own significance. The chapter had also accelerated towards the implementation environments for validating the algorithm performance. In the final section, applications of biomedical imaging analysis using DL for various imaging modalities are presented as state-of-art for the readers. Therefore, several researchers are highly concentrating on DL, based on the efficient solution that mainly encourages using these techniques on biomedical imagining.