Effectiveness of Radiation Therapy for the Treatment of Head and Neck Squamous Cell Carcinoma Open Access
Downloadable ContentDownload PDF
Abstract Head and Neck Squamous Cell Carcinoma (HNSCC) is usually treated with radiation therapy (RT). We wish to study whether the heterogeneity of the tumor affects the clinical outcome after RT. A database is available that contains HNSCC images acquired in various modalities. Each image modality gives different information about the patient; we wish to test whether a combination of modalities will help to assess the likelihood of success of RT. The computerized tomography (CT) scan image gives us anatomical information whereas the positron emission tomography (PET) scan gives us physiological information. The tumor is visible in the pretreatment PET images. PET images have poorer resolutions than CT images, but the PET-CT imaging system ensures that the pre-treatment CT images are registered with the PET images. Thus, using the tumor information in the pre-treatment PET image, we can locate the exact position of the tumor in the pre-treatment CT image. On registering the pre-treatment CT images with the post-treatment CT images, we have the original location of the tumor displayed in the post-treatment CT image, which, in turn will give us the location of the presumably cured tumor in the post-treatment PET image. The effectiveness of the treatment can be judged by the presence, location and size of the tumor at a particular time after the start of the treatment. This involves a study of both the pre- and the posttreatment images. Shifts are observed between the various images, however, because of factors including change in the position of the patient between the imaging sessions. Thus, image registration is important to combine the information obtained by the different imaging modalities and also to correct the shift between the pre-and post-baseline images. The clinical data accompanying the database indicate whether there was a local recurrence of the tumor in the patient after the completion of the treatment. The patients are divided into local recurrent and non-local recurrent categories. The RT Plan, included in the database, is used to indicate the tumor boundaries in the pre-treatment PET images. Once the tumor region has been extracted, various texture measures are computed on it. Those include energy measures, moments of the tumor texture, correlation between the tumor pixels and their neighbors, Fourier measures and others. We also aim to examine whether the heterogeneity of tumor textures can be a factor that predicts the treatability of the tumor by radiation therapy.