The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification (2107.02314v2)

Abstract: The BraTS 2021 challenge celebrates its 10th anniversary and is jointly organized by the Radiological Society of North America (RSNA), the American Society of Neuroradiology (ASNR), and the Medical Image Computing and Computer Assisted Interventions (MICCAI) society. Since its inception, BraTS has been focusing on being a common benchmarking venue for brain glioma segmentation algorithms, with well-curated multi-institutional multi-parametric magnetic resonance imaging (mpMRI) data. Gliomas are the most common primary malignancies of the central nervous system, with varying degrees of aggressiveness and prognosis. The RSNA-ASNR-MICCAI BraTS 2021 challenge targets the evaluation of computational algorithms assessing the same tumor compartmentalization, as well as the underlying tumor's molecular characterization, in pre-operative baseline mpMRI data from 2,040 patients. Specifically, the two tasks that BraTS 2021 focuses on are: a) the segmentation of the histologically distinct brain tumor sub-regions, and b) the classification of the tumor's O[6]-methylguanine-DNA methyltransferase (MGMT) promoter methylation status. The performance evaluation of all participating algorithms in BraTS 2021 will be conducted through the Sage Bionetworks Synapse platform (Task 1) and Kaggle (Task 2), concluding in distributing to the top ranked participants monetary awards of $60,000 collectively.

• The paper presents a comprehensive benchmark for automated brain tumor segmentation using multi-parametric MRI data from 2,040 patients.
• It employs a fusion of advanced segmentation models followed by expert refinement to delineate tumor sub-regions accurately.
• The paper also facilitates MGMT radiogenomic classification using standard metrics like Dice Score and AUC, paving the way for improved neuro-oncology interventions.

Overview of the RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification

The BraTS 2021 challenge represents a decadal milestone for brain tumor segmentation and radiogenomic classification tasks, organized by RSNA, ASNR, and MICCAI. It continues to provide a benchmark for algorithms processing brain glioma segmentation by utilizing multi-institutional multi-parametric Magnetic Resonance Imaging (mpMRI) data from 2,040 patients. Two key tasks constituted the focus of this challenge: segmentation of histologically distinct brain tumor sub-regions and classification of the tumor’s O[6]-methylguanine-DNA methyltransferase (MGMT) promoter methylation status.

Research Motivation

Glioblastoma (GBM) and other gliomas represent the most aggressive primary tumors in the central nervous system. Accurate MRI-based segmentation of tumor sub-regions is critical for planning surgical treatment, image-guided interventions, and therapeutic evaluations. However, manual segmentation is labor-intensive and subjective, highlighting the necessity for automated solutions. Likewise, the determination of MGMT promoter methylation is crucial, given its role as a prognostic factor and predictor for chemotherapy response.

Dataset and Methodological Framework

The BraTS 2021 dataset comprises a comprehensive repository of mpMRI scans featuring T1, post-contrast T1-weighted (T1Gd), T2, and T2-FLAIR volumes, sourced from diverse clinical settings, enhancing the variability and realism of the data. Pre-processing steps, including skull-stripping and stereotactic alignment, ensure homogeneity across the dataset. Notably, the size of the dataset has increased significantly compared to previous years, providing 2,000 glioma cases from several institutionally acquired data.

The task of tumor segmentation required participants to focus on delineating the enhancing tumor (ET), necrotic core (NCR), and peritumoral edema (ED). Initial segmentations were produced via the STAPLE fusion of nnU-Net, DeepScan, and DeepMedic models, followed by expert neuroradiologist refinement.

For radiogenomic classification, MGMT status was derived using varying genomic techniques, including pyrosequencing. The binary classification of methylated and unmethylated tumors was supplied in a privacy-compliant format.

Performance Evaluation

The challenge utilized standard metrics for performance evaluation: Dice Score, Hausdorff Distance for segmentation, and AUC for classification tasks. Rankings were determined based on these metrics, emphasizing accuracy and precision across diverse testing conditions. The comprehensive ranking system allowed robust peer comparison through a permutation-based statistical analysis, ensuring fairness.

Discussion and Future Directions

Despite delivering a robust evaluation framework, the challenge acknowledged limitations such as non-assessed inter-rater reliability and variations in MGMT assessment methodologies across institutions. Future directions propose broader datasets encompassing various brain abnormalities beyond gliomas and incorporating postoperative scan evaluations to assess treatment responses effectively.

The vision extends toward adopting federated learning methodologies to overcome data siloing, promoting enhanced algorithmic robustness. This approach is projected to offer insights for neuro-oncology computational research, enabling more generalized and performance-optimized machine learning models.

Conclusion

The RSNA-ASNR-MICCAI BraTS 2021 challenge lays critical groundwork for computational neuroscience through high-quality, annotated mpMRI datasets. It ensures a continuous evolution in automated medical imaging, permitting enhanced stratification of treatment pathways and diagnostic accuracy in neuro-oncology. The insights garnered from BraTS 2021 are slated to propel subsequent research into more expansive and integrative forms of tumor data analytics.