A simple table lookup method for PET/CT partial volume correction using a point-spread function in diagnosing lymph node metastasis

Objective: We evaluated the partial volume effect in PET/CT images and developed a simple correction method to address this problem. Methods: Six spheres and the background in the phantom were filled with F-18 and we thus obtained 4 different sphere-to-background (SB) ratios. Thirty-nine cervical lymph nodes in 7 patients with papillary thyroid carcinoma (15 malignant and 24 benign) were also examined as a preliminary clinical study. First, we developed recovery coefficient (RC) curves normalized to the maximum counts of the 37-mm sphere. Next, we developed a correction table to determine the true SB ratio using three parameters, including the maximum counts of both the sphere and background and the lesion diameter, by modifying the approximation formula of the RC curves including the point-spread function correction. The full width at half maximum in this formula is estimated with the function of the SB ratio. Results: In the phantom study, a size-dependent underestimation of the radioactivity was observed. The degree of decline of RC was influenced by the SB ratio. In preliminary clinical examination, the difference in the SUV _max between malignant and benign LNs thus became more prominent after the correction. The PV correction slightly improved the diagnostic accuracy from 95 to 100%. Conclusions: We developed a simple table lookup correction method for the partial volume effect of PET/CT. This new method is considered to be clinically useful for the diagnosis of cervical LN metastasis. Further examination with a greater number of subjects is required to corroborate its clinical usefulness.