Spatio-temporal TGV denoising for ASL perfusion imaging.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Arterial spin labeling (ASL) generates perfusion-weighted images by subtracting label and control images.
• ASL images suffer from low signal-to-noise ratio (SNR), necessitating extensive measurements for quality, especially at high resolutions.
• Existing denoising methods often use only the perfusion-weighted image, limiting their effectiveness.

Purpose of the Study:

• To develop a novel spatio-temporal filtering approach for denoising ASL images.
• To improve image quality, quantitative accuracy, and robustness in ASL perfusion imaging.
• To leverage simultaneous information from label and control image pairs for enhanced denoising.

Main Methods:

• Proposed a new denoising method based on total generalized variation (TGV) regularization.
• The approach jointly denoises control and label images by exploiting temporal and spatial similarities.
• Evaluated the method using virtual ground truth data at various SNR levels and high-resolution in-vivo pulsed ASL datasets.

Main Results:

• The proposed TGV-based spatio-temporal filtering significantly improved ASL image quality.
• Demonstrated enhanced quantitative accuracy and superior robustness against outliers compared to existing methods.
• Achieved better performance than seven state-of-the-art denoising techniques.

Conclusions:

• The novel TGV regularization approach effectively denoises ASL data by utilizing both label and control images.
• This method offers a significant advancement in achieving high-quality, accurate, and reliable ASL perfusion imaging.
• The findings suggest a promising new direction for improving quantitative perfusion measurements in clinical and research settings.