Automatic Symmetry Detection From Brain MRI Based on a 2-Channel Convolutional Neural Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Symmetry detection in brain Magnetic Resonance (MR) images is crucial for improving diagnostic accuracy of neurological conditions.
• Existing methods often rely on local image features, which can be complex and less robust.
• Accurate identification of the mid-sagittal plane (MSP) is a key challenge in brain image analysis.

Purpose of the Study:

• To propose an automatic symmetry detection method for 2-D brain MR images using a novel convolutional neural network (CNN) approach.
• To overcome limitations of traditional methods by eliminating the need for local feature detection and matching.
• To enhance the accuracy and efficiency of identifying the ideal mid-sagittal plane (MSP) in brain scans.

Main Methods:

• Development of a 2-channel convolutional neural network (CNN) for automatic symmetry detection in brain MR images.
• Training the CNN on diverse brain image datasets to learn symmetry patterns.
• Utilizing a scoring and ranking scheme with Poisson-sampled brain patches to identify the optimal symmetry axis.

Main Results:

• The proposed CNN-based method demonstrated excellent performance in symmetry detection across synthesized and in vivo MR images.
• Evaluated on 2166 synthesized and 3064 in vivo brain MR images, including healthy and pathological cases.
• Achieved higher accuracy compared to state-of-the-art methods in identifying the mid-sagittal plane (MSP).

Conclusions:

• The developed automatic symmetry detection method offers a robust and accurate approach for brain MR image analysis.
• The CNN-based strategy effectively identifies the ideal mid-sagittal plane (MSP), improving diagnostic potential.
• This method presents significant advantages over traditional techniques, paving the way for enhanced clinical applications.