Support vector machine learning-based cerebral blood flow quantification for arterial spin labeling MRI.

Area of Science:

• Medical Imaging
• Machine Learning
• Neuroscience

Background:

• Arterial Spin Labeling (ASL) MRI is a non-invasive technique for assessing cerebral blood flow (CBF).
• Accurate CBF quantification is crucial for diagnosing and monitoring various neurological conditions.
• Conventional ASL CBF quantification methods face limitations in signal-to-noise ratio (SNR) and image quality.

Purpose of the Study:

• To develop a novel multivariate machine learning classification-based method for ASL CBF quantification.
• To enhance the accuracy and quality of CBF measurements obtained from ASL perfusion MRI.

Main Methods:

• Utilized a machine learning algorithm, Support Vector Machine (SVM), to separate ASL label and control images.
• Extracted perfusion-weighted images from a multivariate SVM classifier, considering all voxels.
• Compared the proposed method against the standard ASL CBF quantification using synthetic and in-vivo data.

Main Results:

• The multivariate machine learning approach significantly improved the spatial signal-to-noise-ratio (SNR) of ASL CBF images.
• Enhanced image appearance and quality of ASL CBF images compared to the conventional univariate method.
• Demonstrated improved performance in both synthetic and in-vivo ASL datasets.

Conclusions:

• Multivariate machine learning-based classification is a valuable tool for improving ASL CBF quantification.
• The developed method offers a significant advancement in non-invasive perfusion imaging.
• This technique holds promise for more accurate neurological assessment using ASL MRI.