Meta-heuristic Techniques to Train Artificial Neural Networks for Medical Image Classification: A Review

Medical imaging has been utilized in various forms in clinical applications for better diagnosis and treatment of diseases. These imaging technologies help in recognizing body's ailing region easily. In addition, it causes no pain to the patient as the interior part of the body can be examined without difficulty. Nowadays, various image processing techniques such as segmentation, registration, classification, restoration, contrast enhancement and many more exist to enhance image quality. Among all these techniques, classification plays an important role in computer-aided diagnosis for easy analysis and interpretation of these images. Image classification not only classifies diseases with high accuracy but also analyses which part of the body is infected. The usage of Neural networks classifier in medical imaging applications has opened new doors or opportunities to researchers stir-ring them to excel in this domain. Moreover, accuracy in clinical practices and the development of more sophisticated equipment are necessary in the medical field for more accurate and quicker decisions. Therefore, keeping this in mind, researchers started using meta-heuristic techniques to classify the methods. This paper provides a brief survey on the role of artificial neural networks in medical image classification, various types of meta-heuristic algorithms applied for optimization purposes, and their hybridization. A comparative analysis showing the effect of applying these algorithms on some classification parameters such as accuracy, sensitivity, and specificity is also provided. From the comparison, it can be observed that the usage of these methods significantly optimizes these parameters leading us to diagnose and treat a number of diseases in their early stage. © 2022 Bentham Science Publishers.