Partial volume corrected image derived input functions for dynamic PET brain studies: methodology and validation for...

Area of Science:

• Nuclear Medicine
• Radiochemistry
• Neuroimaging

Background:

• Quantitative Positron Emission Tomography (PET) studies require arterial input functions (AIFs) for accurate analysis.
• Arterial blood sampling is invasive and limits the clinical applicability of PET.
• Image-derived input functions (IDIFs) offer a non-invasive alternative to arterial sampling.

Purpose of the Study:

• To evaluate the accuracy and applicability of IDIFs derived from dynamic brain PET scans.
• To assess the impact of reconstruction-based partial volume correction (PVC) settings on IDIF accuracy.
• To determine the optimal method for defining arterial regions of interest (ROIs) for IDIF extraction.

Main Methods:

• Dynamic [11C]flumazenil PET data from 10 subjects were analyzed.
• Partial Volume Correction Ordered Subset Expectation Maximization (PVC-OSEM) reconstruction algorithm settings were optimized.
• Various methods for defining arterial ROIs were evaluated to extract IDIFs.
• IDIFs were compared against standard on-line measured arterial input functions (BSIFs).

Main Results:

• Optimized PVC-OSEM (4 iterations, 16 subsets) with a 4.5 mm FWHM resolution kernel yielded the best results.
• Defining the ROI using the four hottest pixels over the carotid arteries was the most effective method.
• Excellent peak Area Under the Curve (AUC) ratios (0.99±0.09) were achieved between IDIFs and BSIFs.
• Volume of Distribution (V(T)) and K1 values derived from IDIFs closely matched those obtained using BSIFs.

Conclusions:

• Reconstruction-based partial volume correction significantly improves the accuracy of image-derived input functions.
• The validated method allows for accurate [11C]flumazenil PET analysis without the need for invasive arterial sampling.
• This approach enhances the clinical utility of quantitative PET imaging.