Physicians at Seoul National University Hospital (SNUH) have discovered that by analyzing microelectrode-specific signals with artificial intelligence during deep-brain stimulation (DBS) as part of treating Parkinson’s disease, they can predict the operation’s result.

If more data and experiences are accumulated, it will be of great help for treating the disease, the hospital said in a news release on Monday.    

The team, led by SNUH Professors Paek Sun-ha, Kim Hee-chan and Sun Suk-kyu and Professor Park Hwang-hyon of Chungnam National University Sejong Hospital, analyzed the microelectrode predictive records of 34 Parkinson’s disease patients who underwent DBS under general anesthesia using a deep-learning method.

From left, Professors Paek Sun-ha, Kim Hee-chan, Park Hwang-hyon and Sun Suk-kyu have analyzed the microelectrode signals to predict deep-brain stimulation success to treat Parkinson’s using the deep learning method.
From left, Professors Paek Sun-ha, Kim Hee-chan, Park Hwang-hyon and Sun Suk-kyu have analyzed the microelectrode signals to predict deep-brain stimulation success to treat Parkinson’s using the deep learning method.

The team analyzed the signals obtained through microelectrodes using the deep learning method and predicted the results. Then they conducted real surgeries, divided patients’ statuses by the degree of recovery, and compared them with the AI prediction. The team conducted the stimulation on both sides of the brain, applying different ratios, assuming that each stimulus would affect the two sides differently, using the AI algorithm's multiple structures. 

AI showed the highest accuracy of 80.21 percent when the ratios were 5:1 and 6:1. The team said it demonstrated a functional structure similar to the real brain nerve’s basal ganglia. 

“The finding will likely become the new paradigm for conducting DBS procedure on Parkinson’s disease patients,” Professor Paek said. 

Professor Kim also said, “The study is a new trial utilizing the deep learning method to predict DBS prognosis. We expect to develop more supportive systems using AI in the future.” 

The study results were published in the latest issue of PLOS ONE. 

 

 

Copyright © KBR Unauthorized reproduction, redistribution prohibited