Deep learning-based identification of causative genes in lissencephaly using 3D-MRI volumetric datasets

Area of Science:

• Neuroscience
• Genetics
• Medical Imaging

Background:

• Lissencephaly is a severe neuronal migration disorder linked to genetic mutations.
• These genetic variations cause distinct alterations in brain structure.
• Accurate genetic identification is crucial for understanding and treating lissencephaly.

Purpose of the Study:

• To identify causative genes for lissencephaly using deep learning on 3D MRI data.
• To classify brain MRI scans based on specific genetic variations.
• To leverage advanced AI for improved diagnostic capabilities in neurodevelopmental disorders.

Main Methods:

• Adapted 3D Convolutional Neural Network (3D-CNN) architectures for 3D MRI analysis.
• Developed a three-class classification system targeting LIS1, DCX, and other genetic variations.
• Utilized a proprietary 3D MRI dataset of lissencephaly patients.

Main Results:

• The 3D-ResNet18 model achieved over 78% mean classification accuracy.
• A decision-making process using prediction probability thresholds enhanced gene precision.
• Average precision improved by 4.67% for DCX and 4.84% for LIS1.

Conclusions:

• Deep learning models show significant potential for genetic identification in lissencephaly.
• The developed method offers improved precision for identifying key causative genes.
• This AI-driven approach can aid in diagnosing and understanding genetic brain malformations.