Explainable AI-driven 3D U-Net framework for accurate and interpretable brain tumor segmentation in clinical MRI

Purpose This study aims to develop an accurate and interpretable deep learning framework for brain tumor segmentation using multimodal MRI. It addresses the limitation of black-box behavior in existing models by integrating explainable AI to improve clinical trust and decision-making. Design/methodology/approach A hybrid framework combining 3D U-Net with Grad-CAM was developed. The model was trained on the BraTS2020 dataset using T1, T2, FLAIR and T1CE modalities. Patch-based training (128 × 128 × 128) and data augmentation were applied to enhance performance and manage computational constraints. Findings The model achieved Dice Similarity Coefficients of 85.4% (enhancing tumor), 78.6% (non-enhancing core) and 82.1% (edema). Grad-CAM visualizations aligned well with expert annotations, demonstrating reliable interpretability alongside competitive segmentation accuracy. Research limitations/implications This study demonstrates that combining deep learning with explainable AI enhances model transparency without compromising performance. It contributes to the growing field of interpretable medical AI and supports further research in trustworthy clinical decision systems. Practical implications The framework can assist radiologists in accurate tumor delineation, surgical planning and treatment monitoring. Explainability improves confidence in AI-assisted diagnosis, facilitating adoption in clinical workflows. Social implications Improved diagnostic accuracy and transparency can enhance patient outcomes, reduce diagnostic errors and promote ethical AI use in healthcare systems. Originality/value This work uniquely integrates 3D volumetric segmentation with Grad-CAM-based interpretability, providing a balanced solution for accuracy, efficiency and clinical trust in brain tumor analysis.

Authors

Institutions

Publication Details

Journal
International Journal of Human Rights in Healthcare
Published
2026-11-20
DOI
https://doi.org/10.1108/ijhrh-04-2026-0015
Primary Topic
Brain Tumor Detection and Classification
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Explainable AI-driven 3D U-Net framework for accurate and interpretable brain tumor segmentation in clinical MRI

Shahnawaz Ahmad, Jiyaul Mustafa, Shahadat Hussain
International Journal of Human Rights in Healthcare
Brain Tumor Detection and Classification
article

Explainable AI-driven 3D U-Net framework for accurate and interpretable brain tumor segmentation in clinical MRI

Shahnawaz Ahmad, Jiyaul Mustafa, Shahadat Hussain
article en

Abstract

Purpose This study aims to develop an accurate and interpretable deep learning framework for brain tumor segmentation using multimodal MRI. It addresses the limitation of black-box behavior in existing models by integrating explainable AI to improve clinical trust and decision-making. Design/methodology/approach A hybrid framework combining 3D U-Net with Grad-CAM was developed. The model was trained on the BraTS2020 dataset using T1, T2, FLAIR and T1CE modalities. Patch-based training (128 × 128 × 128) and data augmentation were applied to enhance performance and manage computational constraints. Findings The model achieved Dice Similarity Coefficients of 85.4% (enhancing tumor), 78.6% (non-enhancing core) and 82.1% (edema). Grad-CAM visualizations aligned well with expert annotations, demonstrating reliable interpretability alongside competitive segmentation accuracy. Research limitations/implications This study demonstrates that combining deep learning with explainable AI enhances model transparency without compromising performance. It contributes to the growing field of interpretable medical AI and supports further research in trustworthy clinical decision systems. Practical implications The framework can assist radiologists in accurate tumor delineation, surgical planning and treatment monitoring. Explainability improves confidence in AI-assisted diagnosis, facilitating adoption in clinical workflows. Social implications Improved diagnostic accuracy and transparency can enhance patient outcomes, reduce diagnostic errors and promote ethical AI use in healthcare systems. Originality/value This work uniquely integrates 3D volumetric segmentation with Grad-CAM-based interpretability, providing a balanced solution for accuracy, efficiency and clinical trust in brain tumor analysis.

International Journal of Human Rights in Healthcare
Bennett University (IN)
Brain Tumor Detection and Classification
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon