Stop! border ahead: Automatic detection of subthalamic exit during deep brain stimulation surgery.

Area of Science:

• Neurosurgery
• Computational Neuroscience
• Biomedical Engineering

Background:

• Accurate subthalamic nucleus (STN) border definition is crucial for effective deep brain stimulation (DBS) in Parkinson's disease.
• Manual STN border demarcation by neurophysiologists is challenging, particularly distinguishing it from the substantia nigra pars reticulata.
• Inaccurate border identification can lead to suboptimal DBS lead placement and reduced clinical efficacy.

Purpose of the Study:

• To develop and validate machine learning algorithms for real-time, high-accuracy discrimination between the STN and substantia nigra pars reticulata.
• To enable automatic and precise identification of the ventral STN border during DBS surgery.

Main Methods:

• Utilized power spectra from microelectrode recordings for classification.
• Employed a support vector machine (SVM) for initial classification, achieving 97.6% consistency with human experts via 10-fold cross-validation.
• Developed a hidden Markov model (HMM) incorporating microelectrode recording features and trajectory history for real-time STN exit classification.

Main Results:

• The SVM achieved high accuracy in differentiating STN from substantia nigra pars reticulata.
• The HMM identified the STN exit with a mean error of 0.04 ± 0.18 mm.
• The HMM demonstrated a 94% reliability in detecting the STN exit within a 1 mm error margin across 73 additional trajectories.

Conclusions:

• Robust, accurate, and automatic real-time electrophysiological detection of the ventral STN border is feasible.
• Machine learning approaches offer a promising solution for improving the precision of DBS lead placement.
• This technology has the potential to enhance clinical outcomes for Parkinson's disease patients undergoing DBS surgery.