(Wednesday, 7/15/2020) 4:30 PM - 5:30 PM [Eastern Time (GMT-4)]
Room: Track 1
Purpose:
To develop a novel method, combining convolutional neural network (CNN) and ensemble learning (EL), to archive high accuracy and efficiency for glioma detection and segmentation using multiparametric MRIs.
Methods:
We proposed an evolutionary feature selection-based hybrid approach for glioma detection and segmentation on 4 MR sequences (Flair, T1, T1ce, and T2). First, we trained a light CNN to detect glioma and mask the suspected region to handle large batch of MRI images. Second, a differential evolution algorithm was employed to search feature space, which composed of 416 radiomics features extracted from 4 sequences of MRIs and 128 high-order features extracted by the CNN, to generate an optimal feature combination for pixel classification. Finally, we trained an EL classifier using the optimal feature combination to segment whole tumor (WT) and its subregions including non-enhancing tumor (NET), peritumoral edema (ED), and enhancing tumor (ET) in the suspected region. Experiments were carried out using the BraTS2017 dataset in which 228 cases were used for training, and 57 cases were used for evaluation.
Results:
The approach achieved a detection accuracy of 98.8% using these four MRI sequences. The Dice coefficients (and standard deviations) were 0.855±0.04,0.846±0.066, and 0.785±0.059 for segmentation of WT (NET+ET+ED), tumor core (TC) (NET+ET), and ET, respectively. The sensitivities and specificities were 0.919±0.076, 0.907±0.092, and 0.879±0.063; 0.99±0.01, 0.99±0.007, and 0.994±0.004 for the WT, TC and ET, respectively. The performances and calculation times were compared with 4 CNN-based state-of-the-art approaches and 3 classic algorithms, our approach yielded a better overall performance with average processing time of 139.5 sec per set of four sequence MRIs.
Conclusion:
We demonstrated a robust and cost-effective segmentation approach for glioma and its subregions on four MR images. The approach could be used for accurate target delineation for patient stratification and/or outcome prediction for glioma patients.
Funding Support, Disclosures, and Conflict of Interest: Support by National Nature Science Foundation of China (61876138, 61203311), Natural Science Basic Research Program of Shaanxi Province of China (2019JM-365), Scientific Research Program Funded by Shaanxi Provincial Education Department of China (17JK0701), and Graduate Innovation Foundation of Xian University of Posts & Telecommunications (CXJJ2017036).
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
