Magia: Robust Automated Image Processing and Kinetic Modeling Toolbox for PET Neuroinformatics

Area of Science:

• Neuroimaging
• Radiochemistry
• Medical Physics

Background:

• Positron emission tomography (PET) data processing often relies on manual delineation of reference regions, introducing significant inter-operator variability.
• This variability can lead to inconsistencies in quantitative outcomes, impacting the reliability of PET studies.

Purpose of the Study:

• To introduce and validate the Magia toolbox for automated processing of brain PET data with minimal user intervention.
• To assess the accuracy and consistency of the Magia toolbox compared to manual processing methods.

Main Methods:

• The Magia toolbox was evaluated using brain PET data from 30 control subjects for four tracers: [¹¹C]carfentanil, [¹¹C]raclopride, [¹¹C]MADAM, and [¹¹C]PiB.
• Reference regions were manually delineated by five operators and automatically generated by Magia.
• Inter-operator variance of manual delineations and differences between manual and automated results (binding potentials, SUVRs) were analyzed.

Main Results:

• Manual reference region delineation showed substantial inter-operator variance, affecting outcome measures.
• Magia-derived reference regions, though anatomically different, yielded outcome measures consistent with the average of manual estimates.
• Magia showed no bias for [¹¹C]carfentanil and [¹¹C]PiB, and a 3-5% increase in binding potentials for [¹¹C]raclopride and [¹¹C]MADAM.

Conclusions:

• The Magia toolbox reliably processes brain PET data, offering consistent results and mitigating inter-operator variance inherent in manual methods.
• The automated approach enhances the efficiency and reproducibility of quantitative PET imaging analysis.