Comparison of a Bayesian estimation algorithm and singular value decomposition algorithms for 80-detector row CT...

Area of Science:

• Neuroradiology
• Medical Imaging Analysis
• Computational Medicine

Background:

• CT perfusion (CTP) postprocessing algorithms vary significantly in quantitative map generation.
• Direct clinical comparisons of Bayesian estimation algorithms with other methods like singular value decomposition (SVD) are limited.

Purpose of the Study:

• To compare the performance of a Bayesian estimation algorithm against standard and reformulated SVD algorithms for acute ischemic stroke assessment.
• To evaluate the accuracy of infarct volume estimation using different CTP algorithms.

Main Methods:

• CTP data from 36 acute ischemic stroke patients were analyzed.
• Standard SVD, reformulated SVD, and Bayesian estimation algorithms were applied.
• Quantitative parameters (CBV, CBF, MTT, TTP, delay) were compared between affected and contralateral sides.
• Agreement between CTP-estimated and diffusion-weighted imaging-derived infarct volumes was assessed.

Main Results:

• Bayesian estimation showed substantial differences in CBF and MTT compared to reformulated SVD.
• Bayesian estimation correlated better with standard SVD for CBF and MTT.
• The Bayesian algorithm demonstrated nearly ideal regression line parameters for infarct volume estimation.

Conclusions:

• The Bayesian estimation algorithm exhibits enhanced performance in acute ischemic stroke assessment.
• It provides better delineation of abnormal perfusion areas and more accurate infarct volume estimation than SVD algorithms.