Effect of Using 2mm Voxels on Observer Performance for PET Lesion Detection

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Radiology

Background:

• Positron emission tomography (PET) images are commonly reconstructed with in-plane pixel sizes of approximately 4mm for cancer imaging.
• Optimizing image reconstruction parameters is crucial for enhancing diagnostic accuracy in oncologic imaging.

Purpose of the Study:

• To evaluate the impact of using smaller pixel sizes on the detection of oncologic lesions in PET imaging.
• To compare the lesion detection performance between PET images reconstructed with ~2mm and ~4mm pixels.

Main Methods:

• Utilized experimental phantom data from the Utah PET Lesion Detection Database, simulating whole-body FDG PET cancer imaging.
• Reconstructed PET images with 2.036mm and 4.073mm pixels using ordered-subsets expectation-maximization (OSEM) with and without point spread function (PSF) modeling and time-of-flight (TOF).
• Assessed detection performance using a channelized non-prewhitened numerical observer and localization receiver operating characteristic (LROC) analysis.

Main Results:

• Images reconstructed with ~2mm pixels demonstrated higher detection performance than those with ~4mm pixels across all evaluated conditions.
• The improvement in detection from smaller pixels exceeded that of PSF modeling and was approximately half the benefit of TOF.
• Human observer studies confirmed the superior detection performance with smaller pixel sizes.

Conclusions:

• Using smaller voxel sizes in PET reconstruction significantly improves tumor detection performance compared to standard sizes.
• The primary trade-offs for improved detection include a four-fold increase in reconstruction time and data storage requirements.
• This study suggests a potential for enhanced oncologic imaging by adopting smaller pixel sizes in PET reconstruction.