Identification for the cortical 3-Hinges folding pattern based on cortical morphological and structural features.

Area of Science:

• Neuroscience and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• The Cortical 3-Hinges Folding Pattern (3-Hinges) is crucial for understanding intelligence, diagnosing neurological diseases, and brain structure differences.
• Identifying 3-Hinges is challenging due to individual morphological variability and reliance on computationally intensive methods like Gyral-net.

Purpose of the Study:

• To develop a novel deep network model for rapid and accurate identification of 3-Hinges using cortical morphological and structural features.
• To investigate the functional and structural roles of 3-Hinges, particularly differences between adolescent males and females.

Main Methods:

• Extraction of morphological and structural features from the cerebral cortex.
• Construction of feature vectors using the K-nearest neighbor algorithm to mitigate overfitting.
• Implementation of a deep U-shaped network with a squeeze-excitation module for channel correlation learning.

Main Results:

• The proposed deep network model demonstrates superior time efficiency compared to existing methods on adolescent and adult MRI datasets.
• Cortical sulcus information is identified as critical for 3-Hinges identification, with cortical thickness, surface area, and volume providing supplementary data.
• Significant structural differences in 3-Hinges were observed between adolescent genders.

Conclusions:

• The developed deep learning approach offers a computationally efficient solution for identifying the Cortical 3-Hinges Folding Pattern.
• The findings highlight the importance of specific cortical features and reveal significant sex-based structural variations in adolescent brain folding patterns.