Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ReFRM3D: A Radiomics-enhanced Fused Residual Multiparametric 3D Network with Multi-Scale Feature Fusion for Glioma Characterization

Md. Abdur Rahman, Mohaimenul Azam Khan Raiaan, Arefin Ittesafun Abian, Yan Zhang, Mirjam Jonkman, Sami Azam

TL;DR

ReFRM3D introduces a radiomics-enhanced fused residual multiparametric 3D network for glioma segmentation and radiomics-based classification on mpMRI. Built on a 3D U-Net backbone, it adds multi-scale feature fusion, hybrid upsampling, and residual skip mechanisms to improve localization and boundary fidelity, while a radiomics-informed classifier augments subtype discrimination. Across BraTS2019–2021, it achieves DSC values in the low-to-mid 90s for WT, ET, and TC, with WT consistently near 94% and classification accuracies exceeding 98–99%, demonstrating strong segmentation and robust radiogenomic-capable classification. The approach also emphasizes computational efficiency via brain-region cropping, voxel standardization, and tumor-contained slice selection, supporting potential clinical deployment in resource-constrained environments.

Abstract

Gliomas are among the most aggressive cancers, characterized by high mortality rates and complex diagnostic processes. Existing studies on glioma diagnosis and classification often describe issues such as high variability in imaging data, inadequate optimization of computational resources, and inefficient segmentation and classification of gliomas. To address these challenges, we propose novel techniques utilizing multi-parametric MRI data to enhance tumor segmentation and classification efficiency. Our work introduces the first-ever radiomics-enhanced fused residual multiparametric 3D network (ReFRM3D) for brain tumor characterization, which is based on a 3D U-Net architecture and features multi-scale feature fusion, hybrid upsampling, and an extended residual skip mechanism. Additionally, we propose a multi-feature tumor marker-based classifier that leverages radiomic features extracted from the segmented regions. Experimental results demonstrate significant improvements in segmentation performance across the BraTS2019, BraTS2020, and BraTS2021 datasets, achieving high Dice Similarity Coefficients (DSC) of 94.04%, 92.68%, and 93.64% for whole tumor (WT), enhancing tumor (ET), and tumor core (TC) respectively in BraTS2019; 94.09%, 92.91%, and 93.84% in BraTS2020; and 93.70%, 90.36%, and 92.13% in BraTS2021.

ReFRM3D: A Radiomics-enhanced Fused Residual Multiparametric 3D Network with Multi-Scale Feature Fusion for Glioma Characterization

TL;DR

ReFRM3D introduces a radiomics-enhanced fused residual multiparametric 3D network for glioma segmentation and radiomics-based classification on mpMRI. Built on a 3D U-Net backbone, it adds multi-scale feature fusion, hybrid upsampling, and residual skip mechanisms to improve localization and boundary fidelity, while a radiomics-informed classifier augments subtype discrimination. Across BraTS2019–2021, it achieves DSC values in the low-to-mid 90s for WT, ET, and TC, with WT consistently near 94% and classification accuracies exceeding 98–99%, demonstrating strong segmentation and robust radiogenomic-capable classification. The approach also emphasizes computational efficiency via brain-region cropping, voxel standardization, and tumor-contained slice selection, supporting potential clinical deployment in resource-constrained environments.

Abstract

Gliomas are among the most aggressive cancers, characterized by high mortality rates and complex diagnostic processes. Existing studies on glioma diagnosis and classification often describe issues such as high variability in imaging data, inadequate optimization of computational resources, and inefficient segmentation and classification of gliomas. To address these challenges, we propose novel techniques utilizing multi-parametric MRI data to enhance tumor segmentation and classification efficiency. Our work introduces the first-ever radiomics-enhanced fused residual multiparametric 3D network (ReFRM3D) for brain tumor characterization, which is based on a 3D U-Net architecture and features multi-scale feature fusion, hybrid upsampling, and an extended residual skip mechanism. Additionally, we propose a multi-feature tumor marker-based classifier that leverages radiomic features extracted from the segmented regions. Experimental results demonstrate significant improvements in segmentation performance across the BraTS2019, BraTS2020, and BraTS2021 datasets, achieving high Dice Similarity Coefficients (DSC) of 94.04%, 92.68%, and 93.64% for whole tumor (WT), enhancing tumor (ET), and tumor core (TC) respectively in BraTS2019; 94.09%, 92.91%, and 93.84% in BraTS2020; and 93.70%, 90.36%, and 92.13% in BraTS2021.
Paper Structure (32 sections, 34 equations, 6 figures, 10 tables, 3 algorithms)

This paper contains 32 sections, 34 equations, 6 figures, 10 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Datasets
  5. Preprocessing
  6. Brain Region Isolation and Cropping
  7. Standardizing Voxel Intensities
  8. Slice Range Selection
  9. Segmentation Model
  10. Base Model
  11. Context Pathway
  12. Multi-scale Feature Fusion
  13. Hybrid Upsampling and Residual Integration
  14. Residual Skip Connection
  15. Aggregation
  16. ...and 17 more sections

Figures (6)

  • Figure 1: Overview of the proposed glioma segmentation framework. The expansion layers mirror the contraction layers in size and filters.
  • Figure 2: Visualization of histograms of original and Z-score normalized intensity distributions with intensity comparison across multiple MRI scans.
  • Figure 3: Visualization of total slice ranges and tumor slice distribution with midpoints: the average tumor start-slice is 46, ending at 113, with a total slice count of approximately 155 per volume.
  • Figure 4: Overview of the proposed classifier incorporating fused maps from segmentation-based and radiomics features.
  • Figure 5: Epoch-wise accuracy and loss for training and validation
  • ...and 1 more figures