Medical Imaging Using Deep Learning

Medical imaging is an essential tool in diagnosis and the development of treatment plans since it provides visual insight to internal human structure with the help of minimally invasive surgery. However, in recent years deep learning has shaken the radiology world, particularly in the interpretation of X-rays by enhancing accuracy, throughput, and scope of tasks. Convolutional Neural Networks (CNNs) are the deep learning algorithms that have been applied with much success in detecting patterns, performing segmentation of regions of interest, and classifying lesion in X-ray images. Contrary to other approaches, these models learn features from images or other input data automatically, and the procedure reduces errors inherent to feature selection while improving diagnostic accuracy. Other improvements like GANs have taken it to the next level where image quality is improved, noise is reduced and additional data is added resulting in higher image resolution for improved clinical assessment. In this paper, the literature review on deep learning for Chest X-ray image analysis is conducted to present the latest architectures and the limitations encountered in medical imaging. Despite such achievements, some challenges such as lack of data, model interpretability, and generalizing on different patients and image equipment are still key challenges. However, some of the issues that arise out of this include ethical issues and regulations, which act as barriers when it comes to the expansion of these systems in the clinical scene. By presenting current approaches and their weaknesses, this paper is intended to give information regarding deep learning to enhance the analysis of medical images in the upcoming years.