Assessment of the accuracy of a Bayesian estimation algorithm for perfusion CT by using a digital phantom

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Perfusion computed tomography (PCT) generates parametric maps crucial for diagnosing various medical conditions.
• A novel Bayesian estimation algorithm has been proposed to enhance the precision of these maps.
• The comparative accuracy of this Bayesian algorithm against established deconvolution methods remains under investigation.

Purpose of the Study:

• To evaluate the accuracy of the Bayesian estimation algorithm for PCT parametric mapping.
• To compare the quantitative performance of the Bayesian algorithm with optimized deconvolution algorithms using a digital phantom.

Main Methods:

• A digital phantom with embedded concentration-time curves for cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (MTT) was utilized.
• Data were analyzed using the Bayesian estimation algorithm and delay-insensitive singular value decomposition (SVD) algorithms from benchmarked software.
• Correlation and agreement between estimated and true quantitative values were assessed.

Main Results:

• All algorithms demonstrated strong correlations with true CBF, CBV, and MTT values (r = 0.91-0.99).
• The Bayesian algorithm exhibited excellent agreement with true values for all parameters (ICC = 0.90-0.99).
• Mean transit time (MTT) values derived from SVD algorithms showed suboptimal agreement (ICC = 0.81-0.82).

Conclusions:

• The Bayesian estimation algorithm provides accurate quantitative values for CBF, CBV, and MTT in PCT.
• The Bayesian algorithm offers superior agreement for MTT compared to delay-insensitive SVD algorithms.
• This study validates the Bayesian algorithm as a precise tool for PCT quantitative analysis.