PET/CT image textures for the recognition of tumors and organs at risk for radiotherapy treatment planning

摘要

Positron emission tomography/computed tomography (PET/CT) images have been used in the radiotherapy treatment planning, especially in the delineations of biological target volumes (BTVs) of tumors. However, it is not possible to accurately and precisely discriminate between tumors and adjacent normal tissues in PET/CT images if the normal tissues with a high PET standard uptake value (SUV) such as brain stem and other brain tissues that are close to the tumors, or if the sub-clinical tumor volumes with a low SUV are encompassed by normal tissues also with a low SUV. CT image has relatively poor soft tissue contrast and many malignant tumors arise from within soft tissue. Therefore there is little distinction between the CT HUumbers of tumors and the surrounding normal tissues. To accurately and precisely distinguish tumors from adjacent normal tissues, and to spare organs at risk (OARs) in radiotherapy treatment planning, we extracted the PET coarseness and busyness, and CT contrast and coarseness respectively from the neighborhood gray-tone-difference matrices of co-registered PET SUV/CT HU images of tumors. We found that PET busyness and contrast can provide more accurate and precise complementary information for the recognition of tumors than PET SUV, while CT coarseness and contrast can offer useful complementary information for the discrimination of organs at risk¹. Therefore, we proposed to delineate the OARs based on CT coarseness, CT contrast and PET busyness by an adaptive 3D volume growing method with two growing stages to best spare the OARs. Moreover, we proposed to delineate the BTVs based on PET SUV, busyness, and contrast by an hierarchical Mumford-Shah Vector Model via a refined ring-volume of interest (VOI) based on the delineated OARs. Five patient studies were assessed and visually inspected by radiation oncologists. The resulting BTVs were more accurate and more precise, and better spared the OARs than our previous BTVs.