TU-F-12A-03: Using 18F-FDG-PET-CT and Deformable Registration During Head-And-Neck Cancer (HNC) Intensity Modulated Radiotherapy (IMRT) to Predict Treatment Response




To evaluate the effect of deformable vs. rigid registration of pre-treatment 18F-FDG-PET-CT to intra-treatment 18F-FDG-PET-CT on different standardized uptake value (SUV) parameters and investigate which parameters correlate best with post-treatment response in patients undergoing IMRT for HNC.


Pre-treatment and intra-treatment PET-CT (after 20Gy) scans were acquired, in addition to a 12 week post-treatment PET-CT to assess treatment response. Primary and lymph node gross tumor volumes (GTV_PRI and GTV_LN) were contoured on the pre-treatment CT. These contours were then mapped to intra-treatment PET images via rigid and deformable registration. Absolute changes from pre- to intra-treatment scans for rigid and deformable registration were extracted for the following parameters: SUV_MAX, SUV_MEAN, SUV_20%, SUV_40%, and SUV_60% (SUV_X% is the minimum SUV to the highest-intensity X% volume).


Thirty-eight patients were evaluated, with 27 available for classification as complete or incomplete response (CR/ICR). The pre-treatment average tumor volumes for the patients were 24.05cm3 for GTV_PRI and 23.4cm3 for GTV_LN. For GTV_PRI, there was no statistically significant difference between rigid vs. deformable registration across all ΔSUV parameters. For GTV_LN contours, all parameters were significantly different except for ΔSUV_MAX. For deformably-registered GTV_PRI, changes in the following metrics were significantly different for CR vs. ICR: SUV_MEAN(p=0.003), SUV_20%(p=0.02), SUV_40%(p=0.02), and SUV_60%(p=0.008). The following cutoff values separated CR from ICR with high sensitivity and specificity: ΔSUV_MEAN=1.49, ΔSUV_20%=2.39, ΔSUV_40%=1.80 and ΔSUV_60%=1.31. Corresponding areas under the Receiver Operating Characteristics curve were 0.90, 0.81, 0.81, and 0.85, respectively.


Rigidly and deformably registered contours yielded statistically similar SUV parameters for GTV_PRI, but not GTV_LN. This implies that neither registration should be solely relied upon for nodal GTVs. Of the four SUV parameters found to be predictive of CR vs. ICR, SUV_MEAN was the strongest. Preliminary results show promise for using intra-treatment 18F-FDG-PET-CT with deformable registration to predict treatment response.