Association Rules Applied to the Detection of the Subthalamic Nucleus for DBS Surgery in Patients with Parkinson's Disease
Deep brain stimulation (DBS) is a therapy for effective treatment in the improvement of motor disorders of Parkinson's disease (PD) based on a pulse generator and stimulation electrodes implanted in the patient and working as a pacemaker of the brain. The most common stimulation zone for PD is the subthalamic nucleus (STN) of cerebral hemispheres. During stereotactic surgery for DBS implantation, signals are obtained through recording microelectrodes (MER) at different depths of the brain. MER are analyzed by neurophysiologists to define the physiological location of the STN involving several hours of visual and acoustic analysis of MER signals during surgery.
The main objective of this work is to obtain an automatic classification algorithm based on class association rules with a good performance and a brief classification time. An algorithm with these characteristics could be used in the future in the design of a support tool that works in real time during a DBS surgery for STN detection.
The algorithm of Classification Based on Associations was applied to generate 5 classification models, from different variables and discretization cut-off points. Statistical comparisons between the performance indexes showed very significant differences (p <0.001) among all the classifiers. In particular, the M5 model presented the best performance indexes (Accuracy of 91.57% and Sensitivity of 91.6 %) through the execution of 10 rules with high lift each, presenting the potential for STN detection in real time.