Related Experiment Videos
Comparative evaluation of MR-based partial-volume correction schemes for PET.
C C Meltzer1, P E Kinahan, P J Greer
1Department of Radiology, University of Pittsburgh, Pennsylvania 15213-2582, USA.
Summary
Two methods for correcting partial-volume effects in PET imaging were compared. The simpler two-compartment method is better for comparative studies, while the three-compartment method offers higher accuracy for absolute measures.
Area of Science:
- Neuroimaging
- Medical Physics
Background:
- Quantitative Positron Emission Tomography (PET) measurements are affected by partial-volume averaging, which limits spatial resolution.
- This averaging occurs among neighboring tissues with differing tracer concentrations, impacting cerebral blood flow, glucose metabolism, and neuroreceptor binding.
Purpose of the Study:
- To compare two Magnetic Resonance (MR)-based approaches for partial-volume correction of PET images.
- To evaluate the impact of various error sources on these correction methods.
- To assess the utility of these methods in studies of aging and neurodegenerative diseases.
Main Methods:
- Simulations and a multicompartment phantom were used to compare a two-compartment and a three-compartment MR-based partial-volume correction method.
- The two-compartment method corrects for cerebrospinal fluid (CSF) effects.
- The three-compartment method corrects for CSF and gray/white matter partial-volume averaging.
Main Results:
- The three-compartment method achieved high accuracy in simulations, with 100% gray matter recovery.
- However, the three-compartment method was more sensitive to errors like misregistration, resolution mismatch, segmentation errors, and white matter heterogeneity.
- The two-compartment method demonstrated better robustness against these errors.
Conclusions:
- The two-compartment approach is recommended for comparative PET studies due to its robustness.
- The three-compartment algorithm is suitable for absolute quantitative PET measures when accuracy is paramount and errors can be minimized.