Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning
Objective Glioblastoma multiforme (GBM) is the most aggressive and prevalent type of brain tumor, with a median survival time of approximately 15 months despite treatment advancements. Determining the O(6)-methylguanine-DNA-methyltransferase (MGMT) promoter status, specifically its methylation, is c...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SAGE Publishing
2025-04-01
|
| Series: | Digital Health |
| Online Access: | https://doi.org/10.1177/20552076251332018 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850272998898008064 |
|---|---|
| author | Muhammad Sohaib Iqbal Usama Ijaz Bajwa Rehan Raza Muhammad Waqas Anwar |
| author_facet | Muhammad Sohaib Iqbal Usama Ijaz Bajwa Rehan Raza Muhammad Waqas Anwar |
| author_sort | Muhammad Sohaib Iqbal |
| collection | DOAJ |
| description | Objective Glioblastoma multiforme (GBM) is the most aggressive and prevalent type of brain tumor, with a median survival time of approximately 15 months despite treatment advancements. Determining the O(6)-methylguanine-DNA-methyltransferase (MGMT) promoter status, specifically its methylation, is crucial for treatment planning as it provides valuable prognostic information and indicates chemosensitivity. However, current methods require invasive tissue sampling and genetic testing, resulting in time-consuming processes. The non-invasive technique of assessing MGMT status in GBM patients may offer valuable insights to neuro-oncologists, aiding in precise treatment and surgical planning. Methods This research study utilizes two benchmark datasets—BraTS2021 brain tumor segmentation dataset and MGMT promoter status classification dataset—and proposes a pipeline of segmentation-based classification of MGMT promoter status utilizing all modalities of magnetic resonance imaging (MRI) scans by stacking them. The pipeline consists of two phases: in the first phase, a 3D Residual U-Net (3D ResU-Net) architecture is used to segment the brain tumor into sub-regions using a stack of MRI modalities. In the second phase, the segmented tumor voxel obtained from the first phase is input into a 3D ResNet10 model to predict MGMT promoter status. Results The segmentation phase of the pipeline achieves promising results with average dice scores of 0.81, 0.84, and 0.80 for tumor core (TC), whole tumor (WT), and enhancing tumor (ET) regions, respectively, on the internal validation set. The classification phase obtains a ROC–AUC score of 0.66 on the internal validation set. Conclusion This pipeline demonstrates the potential of a non-invasive approach to support neuro-oncologists in brain tumor diagnosis and treatment planning. While still at the research stage, it provides insights into tumor sub-regions and MGMT promoter status, highlighting the role of AI-driven methods in assessing molecular data. Future studies and clinical validation are needed to further explore its applicability in real-world clinical settings. |
| format | Article |
| id | doaj-art-2bbccb9934e74ebf84368f656dd2c1ad |
| institution | OA Journals |
| issn | 2055-2076 |
| language | English |
| publishDate | 2025-04-01 |
| publisher | SAGE Publishing |
| record_format | Article |
| series | Digital Health |
| spelling | doaj-art-2bbccb9934e74ebf84368f656dd2c1ad2025-08-20T01:51:38ZengSAGE PublishingDigital Health2055-20762025-04-011110.1177/20552076251332018Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learningMuhammad Sohaib Iqbal0Usama Ijaz Bajwa1Rehan Raza2Muhammad Waqas Anwar3 Department of Computer Science, , Lahore, Pakistan Department of Computer Science, , Lahore, Pakistan School of Information Technology, Murdoch University, Perth, Australia Department of Computer Science, , PakistanObjective Glioblastoma multiforme (GBM) is the most aggressive and prevalent type of brain tumor, with a median survival time of approximately 15 months despite treatment advancements. Determining the O(6)-methylguanine-DNA-methyltransferase (MGMT) promoter status, specifically its methylation, is crucial for treatment planning as it provides valuable prognostic information and indicates chemosensitivity. However, current methods require invasive tissue sampling and genetic testing, resulting in time-consuming processes. The non-invasive technique of assessing MGMT status in GBM patients may offer valuable insights to neuro-oncologists, aiding in precise treatment and surgical planning. Methods This research study utilizes two benchmark datasets—BraTS2021 brain tumor segmentation dataset and MGMT promoter status classification dataset—and proposes a pipeline of segmentation-based classification of MGMT promoter status utilizing all modalities of magnetic resonance imaging (MRI) scans by stacking them. The pipeline consists of two phases: in the first phase, a 3D Residual U-Net (3D ResU-Net) architecture is used to segment the brain tumor into sub-regions using a stack of MRI modalities. In the second phase, the segmented tumor voxel obtained from the first phase is input into a 3D ResNet10 model to predict MGMT promoter status. Results The segmentation phase of the pipeline achieves promising results with average dice scores of 0.81, 0.84, and 0.80 for tumor core (TC), whole tumor (WT), and enhancing tumor (ET) regions, respectively, on the internal validation set. The classification phase obtains a ROC–AUC score of 0.66 on the internal validation set. Conclusion This pipeline demonstrates the potential of a non-invasive approach to support neuro-oncologists in brain tumor diagnosis and treatment planning. While still at the research stage, it provides insights into tumor sub-regions and MGMT promoter status, highlighting the role of AI-driven methods in assessing molecular data. Future studies and clinical validation are needed to further explore its applicability in real-world clinical settings.https://doi.org/10.1177/20552076251332018 |
| spellingShingle | Muhammad Sohaib Iqbal Usama Ijaz Bajwa Rehan Raza Muhammad Waqas Anwar Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning Digital Health |
| title | Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning |
| title_full | Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning |
| title_fullStr | Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning |
| title_full_unstemmed | Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning |
| title_short | Integrated brain tumor segmentation and MGMT promoter methylation status classification from multimodal MRI data using deep learning |
| title_sort | integrated brain tumor segmentation and mgmt promoter methylation status classification from multimodal mri data using deep learning |
| url | https://doi.org/10.1177/20552076251332018 |
| work_keys_str_mv | AT muhammadsohaibiqbal integratedbraintumorsegmentationandmgmtpromotermethylationstatusclassificationfrommultimodalmridatausingdeeplearning AT usamaijazbajwa integratedbraintumorsegmentationandmgmtpromotermethylationstatusclassificationfrommultimodalmridatausingdeeplearning AT rehanraza integratedbraintumorsegmentationandmgmtpromotermethylationstatusclassificationfrommultimodalmridatausingdeeplearning AT muhammadwaqasanwar integratedbraintumorsegmentationandmgmtpromotermethylationstatusclassificationfrommultimodalmridatausingdeeplearning |