The invention discloses an
artificial intelligence-driven
spinal surgery nerve protection closed-loop decision-making method, which comprises the following steps: synchronously acquiring electromyographic signals, image data and micro
blood flow data in real time, constructing a multi-
modal space-time fusion model by adopting
tensor decomposition and ConvLSTM network to predict
injury risk probability, and outputting a
nerve injury risk level in real time. And a neural protection closed-loop decision used for assisting a doctor to make decision reference is generated by taking the historical
database as reference, a
surgical robot intervention control instruction is automatically output based on the
risk level to carry out active intervention, closed-loop updating is carried out on the multi-
modal space-time fusion model by using new multi-
modal data after intervention, and a closed-loop decision is formed. The defects of one-sided monitoring, static modeling and passive
decision making in the prior art are overcome, the
nerve injury risk identification accuracy is larger than or equal to 92%, the
response delay is smaller than or equal to 300 ms, the nerve protection effect of
spinal surgery can be remarkably improved, and real-time accurate data reference is provided for doctors.