Radiomics and Radiogenomics in Neuro-oncology by Hassan Mohy-ud-Din & Saima Rathore

Author:Hassan Mohy-ud-Din & Saima Rathore
Language: eng
Format: epub
ISBN: 9783030401245
Publisher: Springer International Publishing

Assessment of the accuracy of RF models using the DRF or SRF as input features is done in Fig. 2c. DRFs lead to a significantly higher accuracy (p < 0.05), with an average AUC of 89.15% compared to 78.07% using SRF, for predicting the short-term and long-term of survival outcome of rGBM patients. Comparing the AUCs of the predicted groups using the DRF and SRF, Chi-square test showed a significant p value < 0.0001. This result is consistent with the previous finding using the univariate analysis, in which DRF is more relevant than the SRF in predicting the survival. Once again, we applied the Kaplan-Meier estimator and log-rank test on the predicted groups (short-term and long-term survival) obtained by the RF classifier (Fig. 2d). We observe that the patient groups obtained by DRF or SRF have significantly different survival outcomes with p = 1.5 × 10−6, HR = 2.9, CI = 1.82 − 4.7 and p = 6.8 × 10−6, HR = 2.96, CI = 1.86 − 4.69, respectively. To assess the importance of individual features, we combined the DRF and SRF to train the RF classifier model (Fig. 2e). We find that the most predictive features (importance features > 0) are from DRF (i.e., Large Zone/Low Gray Emphasis).

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.