MACHINE LEARNING APPLICATIONS IN MEDICAL IMAGING: THE ADVANCEMENTS AND CHALLENGES OF USING MACHINE LEARNING TO INTERPRET MEDICAL IMAGES
Paper Title: MACHINE LEARNING APPLICATIONS IN MEDICAL IMAGING: THE ADVANCEMENTS AND CHALLENGES OF USING MACHINE LEARNING TO INTERPRET MEDICAL IMAGES
Authors Name: VENKATESWARANAIDU KOLLURI
Download E-Certificate: Download
Author Reg. ID: TIJER_152108
Published Paper Id: TIJER2202004
Published In: Volume 9 Issue 2, February-2022
Abstract: This review paper outlines application of machine learning (ML) in medical imaging, its significance, challenges, and future scope. We begin with a brief description of medical imaging and the reasons that lead to rapid changes in ML technology for analyzing medical imaging data. We then provide the categorization of ML methodologies used in medical imaging. Under supervised learning, we describe its success with quantification tasks in anatomical and functional imaging. Unsupervised learning has been successful in extracting information from complex data, and we expound its potential with multiparametric and molecular imaging [1]. The value proposition of using predictive models and ML to translate radiomics data into biologically relevant imaging biomarkers and potential impact on clinical decision support is discussed under quantitative imaging. We explain the recent emergence of deep learning methods as it correlates with the tremendous successes seen in computer vision tasks and future revolution for automatic analysis and interpretation of complex medical images. The second half of the article discusses the advancements and challenges of using ML to interpret medical images. While ML has benefited the scientific community with tools for better understanding disease phenotypes and observing subtle phenotypic differences, there are growing concerns that it may hinder the development of medical students and young clinicians in acquiring image interpretation skills [1]. A discussion of recent and future ML trends towards translational science and clinical imaging informatics, and the necessity to instill evidence-based validation science and quantitative imaging into these methods to avoid direct clinical implementation of premature or unproven methods. The potential of ML methods to aid radiologists or obviate the need for them is a contentious issue; while ML tools can augment diagnostic and prognostic specificity and sensitivity for improved patient care, it should be viewed as a partnership to improve radiologist efficiency and quality of care, as a replacement of radiologists may risk loss of the patient-doctor relationship and short-sighted oversight of non-radiological patient data. We conclude with the future potentials and limitations of ML tools to truly impact clinical imaging practice and the necessity of close collaboration between imaging scientists and physician clinicians to guide ML tool development towards clinically relevant and high-impact advancements in disease detection, monitoring, and therapeutic response assessment.
Keywords: Medical imaging, automation, machine learning, Convolutional Neural Network (CNN), Computer Aided Design (CAD), disease diagnostic, classification, segmentation, healthcare system
Downloads: 00011
Page No: 17-21
Country: -, -, India
Research Area: Science and Technology
Published Paper URL: https://tijer.org/TIJER/viewpaperforall?paper=TIJER2202004
Published Paper PDF: https://tijer.org/TIJER/papers/TIJER2202004
ISSN:
2349-9249 | IMPACT FACTOR: 8.57 Calculated By Google Scholar| ESTD YEAR: 2014
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 8.57 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: TIJER(IJPublication)
Click Here to Download This Article
Article Preview