A Simulation Toolkit for Testing the Sensitivity and Accuracy of Corticometry Pipelines

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Replicability of neuroimaging findings is a growing concern.
• Neuroimaging pipelines involve complex numerical procedures impacting result accuracy.
• Automated corticometry pipelines require rigorous validation for reliable lesion detection.

Purpose of the Study:

• To propose and validate a simulation method for estimating the sensitivity and specificity of lesion detection in automated corticometry pipelines.
• To compare the performance of different neuroimaging pipeline versions (CIVET, FreeSurfer) using simulated lesions.
• To assess the impact of lesion size, blurring kernel, and thickness metrics on detection accuracy.

Main Methods:

• Simulated artificial brain lesions across various sizes and regions of interest.
• Applied the simulation method to different versions of CIVET and FreeSurfer pipelines.
• Utilized between-subject and within-subject designs with T1-weighted MRIs from ICBM and IBIS datasets.
• Evaluated sensitivity, specificity, and coefficients of variation.

Main Results:

• The simulation method is sensitive to partial volume effects and lesion size.
• Differences in sensitivity and specificity were identified between tested neuroimaging pipelines.
• Performance varied based on lesion characteristics and pipeline parameters.

Conclusions:

• The proposed simulation method effectively reveals differences in lesion detection capabilities of neuroimaging pipelines.
• This method can aid in software development and release workflows for improved neuroimaging research reliability.
• Standardized validation of automated corticometry pipelines is crucial for robust scientific findings.