Session Item

In H&N cancers, MRI are often registered to planning CTs for better target delineation. Multi-parametric MRI such as DWI can be an alternative to PET images. High b value DWI can generate high intensity signal in tumour regions, differentiating tumour from surrounding healthy tissues. This has allowed multi-parametric MRI to improve target delineation for RT planning. Applying semi-automated segmentation for PET and DWI scans may improve target delineation accuracy, which may lead to precise radiotherapy delivery. The aim of this study was to find out if PET/MR can improve target delineation accuracy for RT planning. In this study, we determined the correlation between SUV from PET images and the apparent diffusion coefficient (ADC) from MR images for H&N cancer patients. Subsequently, the similarity of tumour volumes delineated by oncologists in planning CTs and the tumour volumes determined by the use of functional PET/MR images were compared.

12 patients with H&N cancers were recruited. They had PET/MR scan prior to radiotherapy planning. DWI (b0 and b700) images were used to calculate the ADC values then plotted against SUV from PET images to determine the correlation. Rigid and deformable registrations of PET/MR images to planning CTs were performed. Similarity indexes, dice similarity coefficient (DSC) and jaccard similarity coefficient (JSC) were calculated to compare the overlapping of DWI-based target volume (DWITV) and PET-based target volume (PETTV) against targets delineated by treating oncologists. Rigid and deformable registration similarity results were compared. Statistical analysis was performed in SPSS. Kendall’s tau b was used to compute the correlation of SUVmax and ADC non-parametrically. Wilcoxon signed-rank test was used. It was considered as statistically significant if p-value < 0.05.

There was negative correlation between ADC and SUVmax in the GTV group but it was insignificant. In the lymph node (LN) group, there was statistically significant negative correlation between ADC and SUVmax (Kendall’s tau b = -0.4872, p-value = 0.0216). The volumes of DWITV and PETTV were similar with no significant difference in the GTV and LN group. DCE and JCE similarity indexes were significantly different in the LN group for DWITV and PETTV after deformable registration. (DSC: DWITV p = 0.045, PETTV p = 0.005; JSC: DWITV p = 0.038, PETTV p = 0.017).

The use of PET/MR in RT planning, illustrated that DWITV and PETTV had similar volume sizes. DWITV and PETTV had smaller volumes than GTVs defined by treating oncologists. These smaller defined target volumes by DWI and PET images could lead to tighter treatment volumes. Deformable registration of PET/MR was essential in the neck region to provide greater overlapping with LN targets and to achieve accurate target delineation. Target volume delineation could be improved by using multi-parametric PET/MR images in RT planning for head and neck cancers treatment.