Automatic segmentation of the choroid plexuses: Method and validation in controls and patients with multiple...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Neurology

Background:

• Choroid Plexuses (ChP) are vital for cerebrospinal fluid (CSF) production and form the blood-CSF barrier.
• Clinically significant volumetric changes in ChP are observed in neurological disorders such as Alzheimer's, Parkinson's disease, and multiple sclerosis (MS).
• Accurate ChP segmentation is essential for large-scale studies investigating their role in neurological conditions.

Purpose of the Study:

• To develop a novel, automated method for segmenting Choroid Plexuses (ChP) in magnetic resonance imaging (MRI) datasets.
• To ensure the method is user-friendly, requires minimal preprocessing, and has low memory requirements for large-scale applications.

Main Methods:

• A 2-step 3D U-Net architecture was employed for automated ChP segmentation.
• The method was trained and validated on a research cohort of MS patients and healthy subjects.
• A second validation was conducted on a clinical cohort of pre-symptomatic MS patients.

Main Results:

• The automated method achieved an average Dice coefficient of 0.72 ± 0.01 and volume correlation of 0.86 on the research cohort, outperforming FreeSurfer and FastSurfer.
• On the clinical dataset, the method obtained a Dice coefficient of 0.67 ± 0.01, comparable to inter-rater agreement (0.64 ± 0.02), with a volume correlation of 0.84.
• The proposed segmentation tool demonstrated robustness and suitability for both research and clinical MRI datasets.

Conclusions:

• The developed automated 3D U-Net method provides a reliable and efficient tool for Choroid Plexus segmentation.
• This method facilitates large-scale neuroimaging studies by enabling accurate ChP volumetric analysis in various neurological diseases.
• The tool's performance on clinical data suggests its potential for routine use in neurological disorder assessment.