A Bayesian estimation method for cerebral blood flow measurement by area-detector CT perfusion imaging

Area of Science:

• Neurology
• Radiology
• Medical Imaging

Background:

• Cerebral blood flow (CBF) is a critical parameter in assessing cerebrovascular disease.
• CT perfusion (CTP) is a widely used imaging technique for CBF measurement.
• Traditional CTP analysis methods, such as reformulated singular value decomposition (rSVD), may have limitations in accuracy.

Purpose of the Study:

• To compare the efficacy of Bayesian estimation with advanced noise reduction (BEANR) versus rSVD for CBF measurement using CTP.
• To evaluate both CTP methods against N-isopropyl-p-[(123)I]-iodoamphetamine (¹²³I-IMP) single-photon emission computed tomography (SPECT) as a reference standard.
• To assess the performance of BEANR and rSVD in patients with cerebrovascular disease.

Main Methods:

• Thirty-one patients with suspected cerebrovascular disease underwent CTP and SPECT.
• CBF maps were generated using both BEANR and rSVD algorithms.
• CBF indices (ischemic to contralateral) were calculated and compared to SPECT using correlation and Bland-Altman analyses.

Main Results:

• BEANR showed a significant positive correlation with ¹²³I-IMP SPECT (r=0.55, p<0.0001).
• No significant correlation was found between rSVD and SPECT (r=0.15, p>0.05).
• BEANR demonstrated smaller limits of agreement and a significantly different mean difference in CBF index compared to rSVD when evaluated against SPECT.

Conclusions:

• BEANR offers superior potential utility for CBF measurement in CTP compared to rSVD.
• BEANR provides more reliable CBF quantification in patients with cerebrovascular disease.
• The findings suggest BEANR is a promising advancement for CTP analysis in clinical practice.