Objective: To develop artificial intelligence(AI) machine/deep learning techniques for processing/analyzing MER signals of STN which will help the surgical-team (neurologists,neurosurgeons) for deciding optimal location of the target lesion for optimal therapeutic efficacy of DBS such that it is vital to have exact electrode placement.

Background: DBS is a well developed method for implanting electrodes into PD brain for optimal results after STN-DBS. Intraoperative MER and stimulus effects have helped in targeting STN and other subcortical structures. Support Vector Machines(SVMs) are influential computer learning structures, based on statistical theory of learning, capable of resolving taxonomy, regression estimation problems. Currently, SVMs are the objects of much research in PD, for solving pattern recognitions (Ex. characterizing signatures of sub-cortical-structures: substantia-nigra, Zona incerta, Thalamus Nuclei). SVMs proffer improvements over usual learning techniques: size of network is not formed from the outset and abridged level is mathematically assured. In this study, we investigated efficient novel methodology for the taxonomy of MER of sub-cortical-structures with DBS.

Method: 6patients with PD>6 years as per UKPDS brain-bank criteria with normal cognition and good response to L-dopa, H and Y score of <4 included in this study for surgery. Surgery was planned using a CRW frame with MRI protocol using 5 channel Framelink SW. MER was performed in all subjects extending from 10mm + target to 10mm – STN. Final target selection was based on the effects/side-effects of macrostimulation and confirmed by MRI post op. A 2s recording performed with fs 24kHz (Fig1-4), obtained 48000samples. Considering a trajectory of 13records for each  sub cortical structures, final trajectory is made for 52 recording which has 2496000 samples.

Results: Results showed excellent classification circa ~ 99%. The investigation in this study certainly avoids human intervention through subjectivity in confining subcortical structures mainly STN.

Conclusion: Neural activity changes from one structure to another within the brain, the possibility of targeting faults to DBS necessitates the use of MER observation to prove exact aim in surgery, hence use of our methodology can be used for focusing subcortical structures, mainly STN for stimulus.

« Back to MDS Virtual Congress 2020

MDS Abstracts - https://www.mdsabstracts.org/abstract/effectiveness-of-lead-position-with-microelectrode-recording-based-support-vector-machine-for-characterizing-the-sub-cortical-structures-via-deep-brain-stimulus-in-parkinsons-disease/