Automated processing of 15O-water brain PET data for assessment of cerebral ischemia

Sammanfattning

The objective of this study was to develop and validate an automated software for processing dynamic 15O-water positron emission tomography (PET) brain images to assess cerebral ischemia and neurodegenerative diseases. Additionally, it aimed to investigate age and sex dependence of cerebral blood flow. Methods: The software was developed using Python and advanced image processing libraries, incorporating motion correction, downsampling, kinetic modelling, and stereotactic surface projection. The arterial input function was either measured using arterial sampling or using a second scan over the heart. Perfusion parameters, including 𝐾1 (perfusion), 𝑉𝑇 (volume of distribution), and 𝑉𝑎 (arterial blood volume), were computed using a basis function implementation of the single-tissue compartment models and compared with values obtained through non-linear regression. The software was then used to analyse 21 healthy subjects dynamic PET images to assess age- and sex-related differences in cerebral blood flow. Results: A high correlation and agreement between basis functions and non-linear regression methods was found for 𝐾1 (𝑅2 = 0.997), 𝑉𝑇 (𝑅2= 0.986), and 𝑉𝑎 (𝑅2= 0.937). Bland-Altman plots showed good agreement for 𝐾1 (-0.33 % mean difference), 𝑉𝑇 (-2.13 %), and 𝑉𝑎 (-28.78 %). While the basis function method significantly underestimated Va at low values, this had minimal impact on 𝐾1. Mean brain perfusion showed a significant decline with age (p = 0.049). Females exhibited significantly higher median brain perfusion than males (p = 0.003). Conclusion: The automated software accurately estimates perfusion parameters from 15O-water PET brain scans, providing reliable assessments of cerebral blood flow. Despite a significant underestimation of Va, the impact on 𝐾1 estimation is minimal. Age-related declines and sex differences in brain perfusion were consistent with previous studies. This software enables automated analysis of PET-based data for assessment of cerebral ischemia and neurodegenerative diseases evaluations.