Enhancing Cerebral Blood Flow Quantification: A Comprehensive Review of Denoising, Artifact Correction, and Simulation in Arterial Spin Labeling MRI
TL;DR
This review examines conventional image-processing techniques together with modern machine learning and deep learning approaches developed to improve ASL image quality through denoising and enhancement and discusses the major artifacts that affect ASL acquisition.
Abstract
Arterial Spin Labeling (ASL) Magnetic Resonance Imaging (MRI) is a noninvasive imaging technique used to quantify cerebral blood flow (CBF) by using magnetically labeled arterial blood water as an endogenous tracer. Although ASL eliminates the need for exogenous contrast agents, its widespread clinical use is limited by several challenges, including low Signal-to-Noise Ratio (SNR), susceptibility to motion, and various imaging artifacts. To address these limitations, both traditional denoising techniques and Machine Learning (ML)/Deep Learning (DL)-based approaches have been developed to improve the reliability of ASL by reducing noise, correcting artifacts, and enhancing image quality. In addition, the generation of simulated ASL datasets has become an important strategy for training and validating novel methods when sufficient clinical data are unavailable. This review examines conventional image-processing techniques together with modern machine learning and deep learning approaches developed to improve ASL image quality through denoising and enhancement. It also discusses the major artifacts that affect ASL acquisition and summarizes the simulation methodologies used for the development and evaluation of new algorithms.