A Multi-Modular System for the Visualization and Classification of MER Data During Neurostimulation Procedures

This paper proposes an interactive analysis and visualization tool for the accuracy improvement of electrode placement during neurostimulation therapy surgery. During the procedure, the presented system assists the surgeon in the crucial tissue type detection by providing a fused visualization of the current electrode location and the microelectrode recordings (MER). The system processes the MER in real-time and utilizes a convolutional neural network (CNN) to classify the targeted tissue type. In addition to presenting the MER in its raw waveform, the system also offers the visualization of the frequency domain and the result of the neural network. To further assist the decision-making process, additional visualization methods are integrated into the system. Using the pre-operative taken CT and MRI scans, the system offers 3D visualization in the form of direct volume rendering (DVR) and axis-aligned slice views in 2D. Both domains are enriched by the MER readings and the result of the machine learning classifier.