Optimization of PTV estimation in highly infiltrating glioblastoma multiforme with perifocal vasogenic edema

摘要

Delineation of brain neoplasms are tedious if they exhibit homogenous pixel intensities with adjacent morphologies. Glioblastoma Multiforme is a highly infiltrating neoplasm which is proliferative, undifferentiated and non-enhancing. Neither CT nor MRI offers adequate image quality so as to identify the extent of GBM infiltration into perifocal edema. This article discusses the delineation of non-enhancing, undifferentiated, and highly infiltrating proliferative glioblastoma and perifocal edema from proven preoperative TI contrast MR images and estimation of per slice clinical as well as planning target volumes. Segmentation of GBM focus from edema was accomplished by K-Means clustering. Preprocessing include background elimination, restoration with bilateral filtering, enhancement with histogram equalization, and skull stripping. Pre-processing and segmentation tasks were performed in MATLAB®. Validation of the segmentation results, with manual segmentation as ground truth exhibited promising Dice similarity Index with mean of DSI 0.8017 ±0.0385 standard deviation.