Diffusion weighted MRI for tumor delineation in head and neck radiotherapy

Abstract

In radiotherapy treatments of head and neck cancer, ionizing radiation is used to destroy malignant tumor tissue, while sparing the surrounding healthy tissue. For successful delivery of radiotherapy treatments, an accurate definition of the target is essential. This target is defined using threedimensional imaging techniques such as computed tomography (CT), positron emission tomography (PET) and magnetic resonance imaging (MRI). MRI can visualize, amongst others, local differences in the diffusion of water: diffusion weighted MRI (DW-MRI). Tumors have different diffusion characteristics than normal tissue. With DW-MRI these differences can be visualized and used in target volume delineation. In order to use DW-MRI in a radiotherapy setting, geometric accuracy is vital. The most common method to acquire DW-MRI, DW-EPI, is known for its sensitivity to geometric distortions, especially in areas with large magnetic field inhomogeneities. The geometric accuracy of DW-EPI was evaluated in patients with a retrospective analysis of magnetic field maps. It was found that severe geometric distortions are present when using DW-EPI in the head and neck region. Therefore, the use of an alternative acquisition method, DW-SPLICE, was proposed. This method is based on a turbo spin echo sequence, which is used for standard anatomical imaging and has comparable, high geometric accuracy. The method was implemented and demonstrated in patients. Using DW-SPLICE, diffusion weighted images with excellent geometric accuracy were acquired in head and neck cancer patients. Additionally, the DW-SPLICE technique was extended in order to improve fat suppression. Additional images were acquired in order to yield a dataset which was suitable for water fat separation. The results from the water fat separation were applied in order to provide a more robust and homogeneous fat suppression. The diffusion weighted images acquired with DW-SPLICE were used to generate target volume delineations. Using an intensity threshold, an initial volume was segmented on diffusion weighted images. Subsequently these were manually adjusted based on the diffusion coefficients. These delineations were compared with those based on PET. Pathology validation of PET delineations has shown these to be quite accurate. The (semi-)automatic target volume delineations on DW-SPLICE show good correspondence with target volume delineations based on PET, indicating that, for a large part, both techniques indicate the same target for treatment. Using surgical specimens, obtained after total laryngectomy, pathological validation of DW-MRI can be performed. Patients received an MRI exam with DW-MRI prior to surgery. The surgical specimen of the larynx was processed, resulting in a digitally reconstructed threedimensional volume which was matched to the imaging prior to surgery. The initial results from the first patient show that the target volume delineation based on DW-MRI has very good agreement with the tumor defined on pathology. When diffusion weighted images are acquired with good geometrical accuracy, these images can be a valuable addition for target volume delineation in head and neck radiotherapy.