Computer-Extracted Texture Features to Distinguish Cerebral Radionecrosis from Recurrent Brain Tumors on Multiparametric MRI: A Feasibility Study.
Computer-Extracted Texture Features to Distinguish Cerebral Radionecrosis from Recurrent Brain Tumors on Multiparametric MRI: A Feasibility Study.
复制标题
DOI：
10.3174/ajnr.a4931
复制
发表时间：
2016-12
期刊：
影响因子：
--
通讯作者：
Madabhushi A
中科院分区：
文献类型：
--
作者：
Tiwari P;Prasanna P;Wolansky L;Pinho M;Cohen M;Nayate AP;Gupta A;Singh G;Hatanpaa KJ;Sloan A;Rogers L;Madabhushi A
To determine the feasibility of radiomic (computer extracted texture) features in differentiating radiation necrosis (RN) from recurrent brain tumors on routine MRI (Gadolinium (Gd)-T1w, T2w, FLAIR). A retrospective study of brain tumor MRI obtained after 9-months (or later) post-radio-chemotherapy was collected from two institutions. In total, 58 patient studies were analyzed, consisting of a training (N = 43) cohort from one institution and an independent test (N = 15) cohort from another, with surgical histologic findings confirmed by an experienced neuropathologist at the respective institutions. Brain lesions on MRI were manually annotated by an expert neuro-radiologist. A set of radiomic features was extracted for every lesion on each MRI sequence: Gd-T1w, T2w, FLAIR. Feature selection was employed to identify the top 5 most discriminating features for every MRI sequence on the training cohort. These features were then evaluated on the test cohort via a support vector machine (SVM) classifier. The classification performance was compared against diagnostic reads by two expert neuro-radiologists, who had access to the same MRI sequences (Gd-T1w, T2w, and FLAIR) as the classifier. On the training cohort, the area under the receiver operating characteristic curve (AUC) was highest for FLAIR with 0.79, 95% CI [0.77, 0.81] for primary (N =22), and 0.79, 95% CI [0.75, 0.83], for metastatic subgroups (N = 21). Of the 15 studies in the holdout cohort, the SVM classifier identified 12 of 15 studies correctly, while neuro-radiologist 1 diagnosed 7 of 15, and neuro-radiologist 2 diagnosed 8 of 15 studies correctly, respectively. Our preliminary results appear to suggest that radiomic features may provide complementary diagnostic information on routine MRI sequences that may improve distinction of RN from recurrence, both for primary and metastatic brain tumors.
