Central Nervous System

PV QA 2 - Poster Viewing Q&A 2

MO_10_2718 - Point-of-care local failure and overall survival prediction models for meningioma

Monday, October 22
10:45 AM - 12:15 PM
Location: Innovation Hub, Exhibit Hall 3

Point-of-care local failure and overall survival prediction models for meningioma
O. Morin1, W. C. Chen2, J. Villanueva-Meyer3, E. D. Gennatas1, A. Wu1, S. Cha3, S. Magill4, A. Perry5, P. K. Sneed6, M. W. McDermott4, T. D. Solberg1, G. Valdes1, S. E. Braunstein6, and D. Raleigh7; 1University of California San Francisco, Department of Radiation Oncology, San Francisco, CA, 2University of California San Francisco, San Francisco, CA, 3University of California San Francisco, Department of Radiology, San Francisco, CA, 4University of California San Francisco, Department of Neurological Surgery, San Francisco, CA, 5University of California San Francisco, Department of Pathology, San Francisco, CA, 6University of California, San Francisco, San Francisco, CA, 7University of California, San Francisco, Department of Radiation Oncology, San Francisco, CA

Purpose/Objective(s): A subset of meningiomas follow an aggressive clinical course characterized by local recurrence and poor survival. Targeted radiotherapy strategies could be offered to control these agressive tumors. We hypothesized that aggressive meningiomas may be discriminated by distinct imaging characteristics. To test this hypothesis, we investigated point-of-care prediction models for meningioma patients based on clinical presentation, preoperative imaging and postoperative data.

Materials/Methods: We specifically developed a comprehensive database well represented with aggressive meningiomas. We analyzed a database containing clinical, pathologic and radiographic information from 235 patients who underwent surgery for 257 meningiomas from 1990 to 2015. All cases were re-graded according to current diagnostic criteria. Two neuro-radiologists independently annotated 17 radiographic features, and 154 textural and non-textural radiomic features were extracted from 230 preoperative post-contrast 3D SPGR MR images in accordance with the Imaging Biomarker Standardization Initiative. The resulting feature set was reduced to limit feature correlation prior to multivariable modeling. Random forest models were trained and tested using nested resampling. The performance of each model was assessed by calculating mean balanced accuracy (BA) and Area Under the Curve (AUC). Feature importance was extracted to generate insights about the radiographic and radiomic characteristics associated with meningioma grade, local failure and overall survival.

Results: Median follow up was 4.3 years. Meningioma grade was 128 WHO Grade I (50%), 104 Grade II (40%) and 25 Grade III (10%). Models restricted to preoperative information, such as patient demographics and radiographic features, had similar BA (0.61-0.64) and AUC (0.62-0.68) for local failure or overall survival prediction as postoperative models based on meningioma grade and extent of resection. Integrated models incorporating all available demographic, clinical, radiographic, radiomic and pathologic data provided the most accurate estimates of local failure and overall survival (accuracy 0.70, AUC 0.73). Radiomic features alone or in combination with other variables showed moderate and marginal predictive value for grade and local failure, respectively. Of the radiographic features, meningioma diffusion characteristics significantly strengthened prediction of grade and local failure. From the models, we developed decision trees and nomograms to readily calculate the risks of local failure or overall survival from meningioma at different points of care. The most important variables for local failure were tumor size, recurrence status, grade, diffusion score, location, extent of resection and T2 signal.

Conclusion: Models using clinical and quantified imaging features can be used to accurately predict meningioma outcomes, and may be useful for individualizing meningioma radiation treatments.

Author Disclosure: O. Morin: None. W.C. Chen: None. J. Villanueva-Meyer: None. E.D. Gennatas: None. A. Wu: None. S. Cha: None. P.K. Sneed: Honoraria; CareCore National, LLC. Travel Expenses; CareCore National, LLC. Board Member; North American Gamma Knife Consortium. M.W. McDermott: None. T.D. Solberg: Honoraria; BrainLAB. Partnership; Global Radiosurgery Services, LLC. S.E. Braunstein: Advisory Board; Radiation Oncology Questions, LLC. D. Raleigh: None.

Olivier Morin, PhD

Presentation(s):

Send Email for Olivier Morin


Assets

MO_10_2718 - Point-of-care local failure and overall survival prediction models for meningioma



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Point-of-care local failure and overall survival prediction models for meningioma