Machine and Deep Learning Algorithms for Healthcare Applications
K. France, A. Jaya, Doru Eugen Tiliute · 2022
In the healthcare system, medical images are playing a vital role to identify the symptoms of early diseases by using image patterns. In the past few decades, because of increasing advancements in the healthcare systems, it produces a large volume of imaging data with different modalities (MRI, fMRI, tomosynthesis, x-ray, computed tomography, etc.) and different dimensionalities like 2D, 3D, and 4D. So, there is a necessity to develop machine learning (ML) tools to manage these healthcare data. In the medical image diagnosing process, these ML tools are used to automatically identify the different disease patterns that appear from the various modalities of medical images. One of the key challenges in medical image diagnosis using ML tools is representing a medical image in the semantic space, and extracting effective features is the crucial step; it is called the semantic gap. In the past 20 years, there are a lot of enhancements in ML techniques to reduce the semantic gap in the diagnosis of medical images. Deep learning (DL) is one of the predominant techniques, which is extensively used to reduce the semantic gap in medical image diagnosis. DL is subset of Artificial Neural Network (ANN). It has many hidden layers to learn complex patterns with different stages of abstraction. Also, in the medical image diagnosing process, DL algorithms give more accurate result than a radiologist. Particularly, the Convolutional Neural Network (CNN) was developed to study medical images. This paper gives an overview of various DL algorithms used in medical image diagnosis. We aimed to provide the key reach areas, like classification of medical images, segmentation, disease localization, and image retrieval. This will help the researcher to identify the emerging trends, research obstacles, and possible future directions in medical image diagnosis.