Convolutional Neural Networks for Segmenting Cerebellar Fissures from Magnetic Resonance Imaging

Area of Science:

• Neuroimaging
• Medical image analysis
• Deep learning

Background:

• The cerebellum is crucial for motor coordination, and its damage in diseases like spinocerebellar ataxias leads to progressive motor deficits.
• Accurate detection of cerebellar damage aids in disease staging and statistical analysis for treatment planning.
• Manual segmentation of cerebellar structures from MRI is labor-intensive and impractical for large datasets.

Purpose of the Study:

• To propose and evaluate convolutional neural networks (CNNs) for the automated segmentation of cerebellar fissures in brain MRI.
• To develop three distinct CNN models for generating binary masks of fissures, cerebellum with fissures, and cerebellum without fissures.

Main Methods:

• Utilized a convolutional neural network (CNN) architecture for image segmentation.
• Trained three models based on the same CNN architecture to produce specific binary masks.
• Applied the models to brain magnetic resonance imaging (MRI) data.

Main Results:

• The developed CNN models achieved high precision and efficiency in segmenting cerebellar fissures.
• Demonstrated the feasibility of training CNNs for automated cerebellar segmentation.
• Generated accurate binary masks for fissures, cerebellum with fissures, and cerebellum without fissures.

Conclusions:

• CNNs are effective tools for the automated segmentation of cerebellar fissures from MRI.
• This deep learning approach can assist specialists in the diagnosis and characterization of neurodegenerative diseases.
• Automated segmentation offers a more efficient alternative to manual methods for large-scale image analysis.