Click here to


Are you sure ?

Yes, do it No, cancel

MRI Radiomics for Predicting a Poor Prognosis in Patients with GBM

P Borges1, J Lizar1, G Viani2, J Pavoni1*, (1) Department of Physics, Faculty of Philosophy, Sciences and Letters at Ribeirao Preto - University of Sao Paulo,BR, (2) Radiotherapy Department, Ribeirao Preto Medical School Hospital and Clinics, University of Sao Paulo, BR


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Glioblastoma multiforme (GBM) is the most lethal and aggressive brain tumor. Magnetic resonance imaging (MRI) is currently used to diagnose and monitoring it. Quantitative features extracted from these images, such as radiomics features, are being used for correlations with the disease prognosis. This study followed this methodology in a workflow adapted from radiotherapy planning, to assess features related to a poor prognosis among patients with GBM.

Methods: T1 post-contrast MRI from 43 patients were evaluated and 105 radiomics features were extracted. All MRIs were previously rescaled to a common resolution to avoid variation in the results. The Progression Free Survival (PFS) time of all patients was also achieved. These patients were separated into two groups: PFS within 3 months (labeled as 1 for classification) and PFS higher than 3 months (labeled as 0 for classification). The features values of the groups were compared using Shapiro-Wilk and t-Student tests.

Results: Kurtosis was the only significant feature (p<0.05), with different mean values in the two groups. Therefore, a predictive model using a threshold of kurtosis value at 2.7 was created to classify which patient would present a recurrence within 3 months. It presented a global accuracy of 0.7 and a ROC AUC of 0.78.

Conclusion: An insight into the understanding between kurtosis and PSF within 3 months was achieved; this information may be valuable for physicians in the clinical management of the patient.


Image Analysis


IM- MRI : Radiomics

Contact Email