A Semiautomatic Method for Multiple Sclerosis Lesion Segmentation on Dual-Echo MR Imaging: Application in a...

Area of Science:

• Medical Imaging
• Neurology
• Computer-Aided Diagnosis

Background:

• Multiple Sclerosis (MS) lesion segmentation is crucial for research and clinical trials.
• Current manual segmentation methods are time-consuming and prone to variability.
• Automated segmentation can improve efficiency and reliability.

Purpose of the Study:

• To present a multicenter validation of a semiautomatic method for hyperintense MS lesion segmentation.
• To assess the robustness and generalizability of the segmentation technique across different centers and scanners.

Main Methods:

• A region-growing algorithm initiated by expert manual lesion identification.
• A final segmentation-refinement step was incorporated.
• Validation involved 52 patients with relapsing-remitting MS across 6 European centers using dual-echo MRI.

Main Results:

• The method achieved good agreement with manual segmentation (Dice coefficient = 0.62, RMSE = 2 mL).
• Optimization was achieved without the need for a training dataset.
• No significant differences in algorithm performance were observed across different MR scanner manufacturers (P > .05).

Conclusions:

• The semiautomatic segmentation method is robust and does not require center-specific training.
• The technique is suitable for application in clinical settings.
• Adoption can enhance reliability and reduce operator time in MS research and clinical trials.