Image-derived input functions from dynamic ¹⁵O-water PET scans using penalised reconstruction

Area of Science:

• Nuclear Medicine
• Neuroimaging
• Quantitative Imaging

Background:

• Quantitative brain PET scans typically require arterial blood sampling, which is logistically challenging.
• Image-derived input functions (IDIFs) offer a solution but are limited by PET resolution.
• This study addresses the challenge of obtaining accurate IDIFs for dynamic PET imaging.

Purpose of the Study:

• To develop and validate a method for generating accurate image-derived input functions (IDIFs) from dynamic 15O-water PET scans.
• To compare IDIFs with traditional arterial blood-sampled input functions (BSIFs) as ground truth.
• To assess the feasibility of routine clinical use of quantitative cerebral blood flow (CBF) measurements.

Main Methods:

• Employed penalised reconstruction, iterative thresholding, and partial volume correction methods to generate IDIFs.
• Utilized data from 16 subjects with two dynamic 15O-water PET scans (baseline and acetazolamide-administration).
• Compared IDIFs against continuous arterial blood sampling (BSIFs) for validation.

Main Results:

• IDIFs and BSIFs showed good agreement in input curve characteristics (R² values of 0.95 for peaks, 0.70 for tails, 0.76 for peak-to-tail ratios).
• Grey matter cerebral blood flow (CBF) values derived from IDIFs closely matched BSIF-derived CBF, with a mean difference of 2% ± 7.3% coefficient of variation.
• The developed method demonstrated robust performance in generating accurate input functions.

Conclusions:

• A robust IDIF can be generated for dynamic 15O-water PET scans using only the PET images.
• This method eliminates the need for arterial blood sampling, MRI, or complex analytical techniques.
• Routine clinical use of quantitative CBF measurements with 15O-water PET is now feasible.