Developing A Consistent Pet Methodology For Quantitative Assessment Of Flt Pet

Introduction . 3'-deoxy 3'-18-Fluorothymidine (FLT) PET-CT measures cellular proliferation and is being investigated as an surrogate response biomarker in oncology . Standard uptake values (SUV) are the most commonly used quantitative measure in FLT PET. However, there is a lack of consensus on the most robust method of assessing SUV, limiting reliable assessment of treatment response. Progress has been made in FDG PET to standardise methods1, but has not yet been translated to FLT. . Volume of interest (VOI) metrics influence SUVmean (SUVpeak/max to a lesser extent) and therefore quantitative analysis . VOI can be manually drawn or automatically calculated. Merging (the inclusion of multiple areas of separate anatomical uptake in a single VOI) limits the usefulness of automatic methods (Fig 1) . We aimed to investigate 11 methods of assessing SUV using automated and manual VOI and propose a consistent semi-automated technique to calculate SUVmean/peak, which could be used to overcome merging and be used as a less operator-dependent but more reliable and time-efficient assessment of treatment response than manual methods

Methods . 16 patients with head and neck squamous cell carcinoma of the oropharynx (HNSCC) undergoing 6 weeks of radiotherapy treatment were recruited to the FLAIRE study . Serial FLT PET-CT scans were acquired on a Siemens Biograph-40 PET-CT scanner: baseline, week 2 of treatment and 6 weeks after finishing treatment . VOIs were placed over tumour foci using 11 SUV selection metrics (9 automated, 2 manual). The automated methods were based on SUV 3, 4 or 5 above background and 40/50/70% of SUVmax (with two initial thresholds) . Automated VOIs were redefined manually if they merged with neighbouring avid sites . Data on tumour volume and SUVmax/mean/peak were collected, focusing on the most avid lesions on each scan . SUVpeak was defined as the mean value for a 1cm3 sphere centred on voxel of maximum uptake . Automated VOIs (using percentage of SUVmax) which most closely matched manual VOIs were calculated

Results . 16/16 baseline and 13/16 on-treatment scans demonstrated tumour foci with FLT uptake . 38% of lesions on these scans (11/29) required manual contouring for SUV calculation due to merging or low FLT uptake . 145 VOIs on baseline scan provided data on maximum SUV; 129 provided mean & maximum SUV and volume; 112 provided data for all variables . There was a fall in all SUV parameters between baseline and on-treatment scans (Table 1 and Fig 2) . Mean automated threshold of 63% most accurately defined lesions which were manually contoured

Conclusions . For the degree of uptake demonstrated in this study we propose an automated threshold of 63% of SUVmax will provide a reasonable measure of SUV parameters and tumour volume . A semi-automated quantitative technique based on SUVmax could be applied in the assessment of treatment response and warrants further investigation with a larger dataset