A semi-automatic segmentation method for meningioma developed using a variational approach model.

Area of Science:

• Neurosurgery
• Radiology
• Medical Imaging Analysis

Background:

• Meningioma is the most common primary brain tumor.
• Volumetric post-contrast MRI is the gold standard for meningioma volume delineation.
• Manual segmentation of MRI scans is time-consuming.

Purpose of the Study:

• To investigate the utility of a model-based variational approach for meningioma segmentation.
• To assess the accuracy and reliability of an automated mathematical model for meningioma volume calculation.

Main Methods:

• A mathematical model was developed for meningioma segmentation.
• The model's performance was evaluated against manual segmentation by a neuroradiologist.
• Segmentation accuracy was quantified using Sørensen-Dice coefficient (DICE) and JACCARD index.
• The model was validated on a separate public dataset of 708 meningioma slices.

Main Results:

• The mathematical model successfully segmented 48 out of 49 meningioma cases.
• Median volumes from manual and model-based segmentation were comparable (19.0 cm³ vs. 16.9 cm³).
• High accuracy was achieved with mean DICE score of 0.90 and JACCARD index of 0.82 on both datasets.

Conclusions:

• The proposed mathematical model provides accurate meningioma volume segmentation.
• This automated method can potentially reduce the manual workload for neuroradiologists.
• The model shows promise for efficient analysis of contrast-enhanced volumetric MRI in meningioma diagnosis.