DeLTA-BIT: an open-source probabilistic tractography-based deep learning framework for thalamic targeting in...

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• In-vivo tractography is crucial for neurosurgical planning and understanding brain connectivity.
• Current tractography methods are time-consuming and not physician-friendly.
• There's a growing need for patient-specific targeting in functional neurological disorders.

Purpose of the Study:

• To develop a novel, open-source deep learning framework for fast and accurate brain target prediction.
• To introduce DeLTA-BIT (Deep-learning Local TrActography for BraIn Targeting) for probabilistic tractography-based targeting.
• To predict the location of the Ventral Intermediate Nucleus (VIM) of the thalamus using deep learning.

Main Methods:

• A convolutional neural network (CNN) was developed within the DeLTA-BIT framework.
• The CNN was trained on the Human Connectome Project (HCP) dataset.
• Model performance was validated on both HCP (internal) and clinical (external) datasets using T1 images.

Main Results:

• DeLTA-BIT achieved good accuracy in predicting the VIM region on internal validation (mean DSC = 0.62).
• The framework processed data rapidly, taking only a fraction of a second per subject.
• External validation on clinical data showed comparable performance to atlas-based methods but with significantly reduced processing time.

Conclusions:

• DeLTA-BIT offers a fast and accurate solution for patient-specific brain targeting in neurosurgery.
• The deep learning approach streamlines the use of tractography for procedures like deep brain stimulation.
• This framework has the potential to improve the efficiency and precision of neurosurgical interventions.