AI-Assisted Brain Tumor MRI Analysis
A research-oriented platform for automated segmentation and quantitative analysis of tumor regions from multi-sequence brain MRI.
Upload MRI Study
Select all MRI volumes belonging to one study. The system automatically identifies T1, contrast-enhanced T1, T2 and FLAIR from supported filename conventions.
Click here or drag and drop the MRI NIfTI files.
Supported formats: .nii and .nii.gz
Analysis Results
Whole Tumor (WT)
Tumor Core (TC)
Enhancing Tumor (ET)
Experimental Glioma Grade-Risk Estimate
Downloads
Interactive MRI Segmentation Viewer
Switch between MRI sequences, change the viewing plane and adjust the tumor-overlay opacity while inspecting the same predicted segmentation.
Uploaded MRI volumes and generated results are temporarily retained for up to 6 hours and can be deleted immediately after use.
How the System Works
The platform combines automated MRI study recognition, deep-learning segmentation, quantitative tumor analysis and interactive visualization.
MRI Acquisition
Four complementary MRI sequences are supplied for the same study.
Study Recognition
The platform identifies and validates T1, T1ce, T2 and FLAIR volumes.
AI Segmentation
The MRI Deployment Model predicts tumor-region masks from four MRI channels.
Analysis
Tumor volumes, masks, a PDF report and interactive overlays are produced.
MRI Information Used by the Model
T1
Provides detailed anatomical information about brain structure.
T1ce
Contrast-enhanced T1 can highlight actively enhancing tumor tissue.
T2
Provides sensitivity to fluid-rich tissue and abnormalities.
FLAIR
Suppresses normal fluid signal and highlights many abnormal brain regions.
End-to-End Processing Pipeline
Predicted Tumor Regions
Whole Tumor - WT
The broadest predicted tumor-associated region.
Tumor Core - TC
The predicted central tumor component within the whole-tumor region.
Enhancing Tumor - ET
The predicted contrast-enhancing tumor component.
Model Performance
Performance was measured on a held-out reference evaluation cohort and separately on an external cohort to assess generalization.
Reference Test Performance
External Generalization Performance
Dice measures spatial overlap between predicted and reference masks. These cohort-level values are not probabilities that an individual patient has cancer. Dataset provenance and experimental protocols remain documented in the research study.
How the AI Model Was Developed
Additional multimodal information is used during research training while the deployed system remains MRI-only.
Experimental controls showed no advantage over matched supervised fine-tuning on the reference test cohort, while evidence of improved external robustness was observed, particularly for enhancing-tumor segmentation.
Current System Capabilities
What the System Can Do
- Automatically organize supported multi-sequence MRI studies.
- Segment Whole Tumor, Tumor Core and Enhancing Tumor.
- Estimate tumor-region volumes.
- Provide an experimental glioma lower-grade/higher-grade MRI-pattern estimate when a full supported study is available.
- Display MRI and segmentation overlays.
- Export masks and a PDF analysis report.
Caution
AI-generated results should be reviewed and interpreted by a qualified healthcare professional before any clinical decision is made. The system is intended to support analysis and should not replace professional medical judgment.
