[3D volume and SUV analysis of oncological PET studies: a voxel-based image processing tool with NSCLC as example].

AIM

The standardized uptake value (SUV) of 18FDG-PET is an important parameter for therapy monitoring and prognosis of malignant lesions. SUV determination requires delineating the respective volume of interest against surrounding tissue. The present study proposes an automatic image segmentation algorithm for lesion volume and FDG uptake quantitation.

METHODS

A region growing-based algorithm was developed, which goes through the following steps: 1. Definition of a starting point by the user. 2. Automatic determination of maximum uptake within the lesion. 3. Calculating a threshold value as percentage of maximum. 4. Automatic 3D lesion segmentation. 5. Quantitation of lesion volume and SUV. The procedure was developed using CTI CAPP and ECAT 7.2 software. Validation was done by phatom studies (Jaszczak phantom, various "lesion" sizes and contrasts) and on studies of NSCLC patients, who underwent clinical CT and FDG-PET scanning.

RESULTS

Phantom studies demonstrated a mean error of 3.5% for volume quantification using a threshold of 41% for contrast ratios >or=5 : 1 and sphere volumes >5 ml. Comparison between CT- and PET-based volumetry showed a high correlation of both methods (r = 0.98) for lesions with homogeneous FDG uptake. Radioactivity concentrations were underestimated by on average -41%. Employing an empirical threshold of 50% for SUV determination, the underestimation decreased to on average -34%.

CONCLUSIONS

The algorithm facilitates an easy and reproducible SUV quantification and volume assessment of PET lesions in clinical practice. It was validated using NSCLC patient data and should also be applicable to other tumour entities.