Deep Learning-Based Deep Brain Stimulation Targeting and Clinical Applications.

Area of Science:

• Neurosurgery
• Artificial Intelligence
• Medical Imaging

Background:

• Deep brain stimulation (DBS) is a crucial treatment for neurological disorders.
• Accurate surgical planning is essential for effective DBS placement.
• Current targeting methods may face limitations with anatomical variability.

Purpose of the Study:

• To evaluate deep learning-based image-guided surgical planning for DBS.
• To develop and clinically apply a semantic segmentation method for DBS targeting.

Main Methods:

• Utilized T2*-weighted MRI images from 102 patients for training and validation.
• Employed a fully convolutional neural network (FCN-VGG-16) for semantic segmentation of subthalamic and red nuclei.
• Augmented image contrast and incorporated manual ground truth for network training.

Main Results:

• Achieved high accuracy (mean accuracy 0.904, mean intersection over union 0.813) in semantic segmentation.
• Demonstrated adaptability to significant anatomical variations in target structures.
• Successfully applied the method in two patients, showing clinical improvement without complications.

Conclusions:

• Deep learning-based semantic segmentation offers high accuracy, potentially exceeding previous methods.
• This AI approach enables precise DBS targeting and clinical application, accommodating anatomical differences.
• The method holds promise for enhancing surgical planning and outcomes in DBS procedures.