Improved residue function and reduced flow dependence in MR perfusion using least-absolute-deviation regularization.

Area of Science:

• Medical Imaging
• Neuroimaging
• Biophysics

Background:

• Cerebral blood flow (CBF) estimation using singular value decomposition (SVD) in perfusion-weighted imaging often underestimates CBF, particularly at high flow rates.
• Accurate CBF quantification is crucial for diagnosing and monitoring various neurological conditions.

Purpose of the Study:

• To develop and validate a model-independent, delay-invariant deconvolution technique using least-absolute-deviation (LAD) regularization to enhance CBF estimation accuracy.
• To compare the performance of the LAD method against standard SVD (sSVD) and reformulated SVD (rSVD) techniques.

Main Methods:

• Computer simulations were conducted to evaluate CBF estimation accuracy across a range of signal-to-noise ratios (20-400), cerebral blood volumes (1%-10%), and CBF values (2.5-176.5 mL/100 g/min).
• The LAD deconvolution technique was implemented and compared with sSVD and rSVD methods.
• Initial clinical validation was performed on six patient cases.

Main Results:

• The LAD method demonstrated improved CBF estimation accuracy by up to 32% in gray matter and 23% in white matter compared to rSVD and sSVD.
• LAD reduced the systematic bias associated with SNR variations and provided more accurate and reproducible residue function calculations.
• Clinical cases confirmed the superiority of the LAD method over existing SVD techniques.

Conclusions:

• The developed LAD deconvolution technique offers a significant improvement in the accuracy and reproducibility of cerebral blood flow estimation from perfusion-weighted imaging.
• This model-independent approach overcomes limitations of SVD methods, particularly in scenarios with high flow rates and varying SNR.
• The LAD method shows promise for enhanced clinical assessment of cerebrovascular health.