Deep learning models for targeting in neurosurgical treatments with transcranial MR-guided Focused Ultrasound Surgery (tcMRgFUS)

In recent years, in vivo tractography has become essential in neuroscience, including noninvasive brain connectivity studies and presurgical planning. There is growing interest in diffusion tractography for target identification in functional neurological disorders, enabling a more tailored approach. This technique is widely used for established neurosurgical treatments like Deep Brain Stimulation (DBS) and is powerful also for newer methods such as trans-cranial Magnetic-Resonance–guided Focused Ultrasound Surgery (tcMRgFUS). Tractography provides more accurate, patient-specific information compared to stereotactic atlases, but it is not very user-friendly and requires hours for MRI data processing. This study presents the development of a deep learning framework for rapid target predictions using probabilistic tractography. It utilizes a Convolutional Neural Network (CNN) to predict the location of the Ventral Intermediate Nucleus of the thalamus (VIM). Trained on Human Connectome Project datasets, the CNN demonstrated strong predictive capability for the VIM region using only T1w images, achieving results in fractions of a second per subject, allowing real-time use during treatment.