|
KMID : 1144720220260040246
|
|
Investigative Magnetic Resonance Imaging
2022 Volume.26 No. 4 p.246 ~ p.255
|
|
|
Deep Learning Applications in Perfusion MRI: Recent Advances and Current Challenges
|
|
Choi Kyu-Sung
|
|
|
Abstract
|
|
|
|
Deep learning has shown its feasibility for applications in medical imaging. Deep learning-based methods are also rapidly being applied in a wide range of areas to replace traditional model-based methods, showing remarkable improvements in several MR image processing areas such as image reconstruction, image contrast conversion, and image quality improvement. With improvement of perfusion MRI techniques, various clinical applications have been also researched, which have improved tracer-kinetic modeling, and vice versa. Representatively, increased vascularity due to tumor angiogenesis, altered permeability due to blood-brain barrier breakdown, and quantifiable absolute cerebral blood perfusion can be imaged via perfusion MRI techniques. As a result, a large number of retrospective and prospective studies have proven that perfusion MRI can be used clinically to investigate various diseases (ranging from brain tumor, stroke, and migraine to neurodegenerative diseases such as Alzheimer’s disease) and various organs including brain, breast, prostate, and pancreas. However, tracer-kinetic model-based processing of parametric maps requires many physical assumptions, which is time-consuming with limitations in clinical applications. With current rapid development of deep learning technologies, the importance of the role of deep learning in MR imaging research appears to be increasing. Specifically for processing perfusion MRI, deep learning has its merit as an end-to-end model-free approach, resulting in markedly reduced processing time compared to conventional iterative methods thanks to its nearly instantaneous inference despite long training time. In this review, first, basic principles of MR physics in perfusion MRI are described. Recent progress and current challenges in both technical improvement and clinical applications of perfusion MRI using deep learning techniques are then summarized.
|
|
|
KEYWORD
|
|
|
Deep learning, Machine learning, Neural network, Dynamic contrast enhanced MRI, Dynamic susceptibility contrast MRI, Arterial spin labeling
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
  
|
|