Validation of magnetic resonance imaging-based automatic high-grade glioma segmentation accuracy via 11C-methionine positron emission tomography.

Validation of magnetic resonance imaging-based automatic high-grade glioma segmentation accuracy via 11C-methionine positron emission tomography. Oncol Lett. 2019 Oct;18(4):4074-4081 Authors: Ozaki T, Kinoshita M, Arita H, Kagawa N, Fujimoto Y, Kanemura Y, Sakai M, Watanabe Y, Nakanishi K, Shimosegawa E, Hatazawa J, Kishima H Abstract Brain Tumor Image Analysis (BraTumIA) is a fully automated segmentation tool dedicated to detecting brain tumors imaged by magnetic resonance imaging (MRI). BraTumIA has recently been applied to several clinical investigations; however, the validity of this novel method has not yet been fully examined. The present study was conducted to validate the quality of tumor segmentation with BraTumIA in comparison with results from 11C-methionine positron emission tomography (MET-PET). A total of 45 consecutive newly diagnosed high-grade gliomas imaged by MRI and MET-PET were analyzed. Automatic tumor segmentation was conducted by BraTumIA and the resulting segmentation images were registered to MET-PET. Three-dimensional conformal association between these two modalities was calculated, considering MET-PET as the gold standard. High underestimation and overestimation errors were observed in tumor segmentation calculated by BraTumIA compared with MET-PET. Furthermore, when the tumor/normal ratio threshold was set at 1.3 from MET-PET, the BraTumIA false-positive fraction was ~0.4 and the false-negative fraction ...
Source: Oncology Letters - Category: Cancer & Oncology Tags: Oncol Lett Source Type: research