Master of Science (MS)
Physics and Astronomy
Positron emission tomography (PET) and computed tomography (CT) together are a powerful diagnostic tool, but imperfect image quality contributes to false positive and false negative diagnoses by an observer despite experience and training. This work investigated a PET standard uptake value (SUV) correction scheme, based on partial volume effect (PVE), on the classification of lesions as benign or malignant in PET/CT images. The correction scheme comprised several steps. The observer drew a region of interest (ROI) around the lesion using the CT dataset. The ROI was blurred with the assumed point spread function (PSF) of the PET scanner then re-sampled to the PET voxel size. The magnitude of the ROI-based PVE was used as a scaling factor to correct the lesion’s tumor-to-background ratio (TBR), which was used as a surrogate for SUV in the PET images of the phantom. Computer simulations showed that the accuracy of the correction depends strongly on the accuracy of the ROI drawn on the CT images, especially for small lesions. Correction accuracy was affected slightly by mismatch of the simulation PSF to the actual scanner PSF. A receiver operating characteristic (ROC) study, using phantom data, was performed to evaluate the effect of the correction scheme on diagnostic performance. The correction scheme significantly increased sensitivity and slightly increased accuracy for all acquisition and reconstruction modes at the cost of a small decrease in specificity. Corrected TBRs more accurately represented actual TBRs than uncorrected TBRs. The observer study also found that, when using PET data alone, 3D ordered subset expectation maximization (OSEM) outperformed 3D filtered back-projection (FBP), 2D OSEM, and 2D FBP in terms of sensitivity, specificity, and area-under-the-ROC-curve values. However, when PET data was displayed with correlated CT data, with and without PVE correction, no combination of reconstruction algorithm and acquisition mode outperformed any other.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Morrow, Andrew Nicholas, "PET/CT detectability and classification of simulated pulmonary nodules using an SUV correction scheme" (2008). LSU Master's Theses. 904.