The use of deformable image registration to integrate diagnostic MRI into the radiotherapy planning pathway for head and neck cancer.

Abstract

To assess the accuracy of gross tumour volume (GTV) delineation for head and neck squamous cell carcinoma (HNSCC) using a diagnostic position MRI (MRI-D) deformably registered to the planning CT (pCT), by comparison with a dedicated planning position MRI (MRI-RT). Fourteen patients with HNSCC underwent a T1-weighted MRI-D and MRI-RT. A reference GTV was defined as that delineated on MRI-RT rigidly registered to pCT. GTVs were delineated on: MRI-D and then registered to pCT by deformable image registration over the whole image (DIR-whole); MRI-D and then registered to pCT with rigid registration over a region of interest defined as GTV+3cm (Rigid-ROI); and on the pCT alone. These were compared using positional metrics to the reference. GTVs delineated on MRI-D followed by DIR-whole were significantly more accurate than those delineated on CT alone. The mean Dice Similarity Coefficient was 0.6 and 0.72 for pCT and DIR-whole respectively. Use of MRI-D with Rigid-ROI provided no advantage over CT-only delineation. Contouring on MRI-RT rigidly registered to pCT should be considered as the gold standard for HNSCC. In radiotherapy centres lacking a dedicated MRI-RT, the use of an MRI-D with DIR-whole offers a significant advantage for the accuracy of GTV delineation over contouring on pCT alone.