Abstract #2225

Tumor progression prediction in high grade glioma using multimodal image analysis and random forest machine learning

Charlotte Debus^1,2,3,4, Maximilian Knoll^1,2,3,4, Sebastian Adeberg^3,4, Stefan Rieken^3,4, Jürgen Debus^1,2,3,4, and Amir Abdollahi^1,2,3,4

¹German Cancer Consortium (DKTK), Heidelberg, Germany, ²Translational Radiation Oncology, National Center for Tumor Diseases (NCT), German Cancer Research Center (DKFZ), Heidelberg, Germany, ³Heidelberg Institute of Radiation Oncology (HIRO), National Center for Radiation Research in Oncology (NCRO), Heidelberg, Germany, ⁴Heidelberg Ion-Beam Therapy Center (HIT), Department of Radiation Oncology, Heidelberg University Hospital, Heidelberg, Germany

Tumor delineation in radiotherapy planning of high grade glioma is challenging due to infiltrative growth patterns and physiological tumor heterogeneity. We used random forest machine learning to classify tissue types and predict tumor progression based on parameters derived from multi-modal functional and metabolic imaging. In an integrative approach, eight patients with recurrent high grade glioma were investigated retrospectively, and the resulting predicted tumor volumes were compared to standard T₁ weighted contrast-enhanced MRI based segmentations. Predictions of tumor tissue could identify original tumor volumes well and yielded promising results with respect to tumor progression in terms of sensitivity and specificity.

This abstract and the presentation materials are available to members only; a login is required.

Join Here