Evaluating the role of partial-volume correction in predicting survival in head and neck cancer

Background: Semi-quantitative analysis of ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG-PET/CT) can be hampered by the partial volume effect (PVE), especially for small sized lesions. The role of PVE in the assessment of head and neck cancer (HNC) survival has not been investigated in any previous study.

Purpose: To examine if applying partial volume correction (PVC) can improve the accuracy of ¹⁸F-FDG-PET/CT in predicting the survival of HNC patients.

Methods: The pre-treatment ¹⁸F-FDG-PET/CT of fifty-four HNC patients with squamous cell carcinoma (47 males, 7 females, median aged 59.5 years, range: 37.2-78.6) were retrospectively assessed. All of patients received linear accelerator (LINAC) for radiation therapy in their course of treatment. A blinded observer performed image analysis of the PET data. Using ROVER software, an adaptive contrast-oriented thresholding algorithm was used to segment lesions on the FDG-PET images. PVC was applied to the PET metrics (ROVER, ABX GmbH, Radeberg, Germany). SUVmean, total lesion glycolysis (TLG), partial volume-corrected SUVmean (pvcSUVmean) and pvcTLG were calculated. Receiver operating characteristic (ROC) was used to determine the area under curve (AUC). Univariate Cox regression analysis was performed to identify the factors associated with progression-free survival (PFS). PFS was defined from the date of start of treatment until disease progression or death.

Results: Applying PVC was associated with a significant increase in the PET metrics (from 7.9± 4.4 to 11.9± 6.9 for SUVmean and from 182.9 ± 157.6 to 264.8 ± 219.5 for TLG; p< .001). AUCs were .425 (95% CI: .266-.585) for SUVmean and .415 (.257-.574) for pvcSUVmean. Moreover, AUC was .592 (.432-.752) for the volumetric PET metric, TLG, before applying PVC and was .590 (.431-.749) after applying PVC. In a univariate Cox regression analysis, both TLG and pvcTLG were prognostic factors for PFS (HR: 1.002, 95% CI: 1-1.005; p= .04 and HR: 1.002, 95% CI: 1-1.004; p= .03, respectively). The results for SUVmean and pvcSUVmean were not significant (p> .05).

Conclusions: This is the first study that assessed the prognostic significance of PVC in head and neck cancers. Although we did not observer a significant improvement for predicting PFS after applying partial volume correction to PET metrics, pvcTLG showed to be a reliable prognostic factor for evaluation of survival in the patients. We further suggest conducting more studies assessing small lesions (less than 1 centimeter), since these lesions are the most susceptible to PVE.

• Download figure
• Open in new tab
• Download powerpoint