Multi-Output Decision Trees for Lesion Segmentation in Multiple Sclerosis

Area of Science:

• Neurology
• Medical Imaging
• Computer Science

Background:

• Multiple Sclerosis (MS) is a central nervous system disease damaging the myelin sheath, leading to lesions in the brain and spinal cord.
• Magnetic Resonance (MR) imaging (MRI) is crucial for MS diagnosis and progression monitoring, with lesion volume and number being key indicators.
• Manual lesion segmentation in MRI is time-consuming and requires specialized expertise.

Purpose of the Study:

• To develop and evaluate an automated algorithm for segmenting Multiple Sclerosis (MS) lesions in MR images.
• To improve the efficiency and accuracy of MS lesion detection and delineation.
• To provide a tool for better diagnosis and tracking of MS progression.

Main Methods:

• Utilized multi-output decision trees for automated lesion segmentation in MR images.
• Algorithm evaluated on the MICCAI 2008 MS Lesion Segmentation Challenge training dataset (20 subjects).
• Performance also assessed on an in-house dataset (49 subjects).

Main Results:

• The automated algorithm demonstrated improved results compared to state-of-the-art methods on the MICCAI dataset.
• On an in-house dataset, the algorithm achieved a true positive rate of 0.41 and a positive predictive value of 0.36.
• The proposed method offers a more efficient alternative to manual lesion segmentation.

Conclusions:

• The developed automated algorithm effectively segments MS lesions in MR images.
• This approach has the potential to enhance the diagnosis and monitoring of Multiple Sclerosis.
• Further validation and refinement could lead to widespread clinical adoption.