Development of automatic diagnostic support system for peripheral nerve lesions in upper limbs using hybrid neural network

摘要

Though short-latency somatosensory evoked potentials (SSEPs) recorded at Erb's point are used as physiological test of peripheral nerve conduction disorder, thoracic outlet syndrome (TOS) cervical radiculopathy (CR) and myelopathy et al., it is difficult for professional docter to diagnose these patints only by SSEPs test. In the present study, we attended to N9 potentials in the SSEPs, and tried to distinguish patients with two diseases from normal subjects using a hybrid neural network. We selected two kinds of input data, thus the raw and the secondary differential waveforms during about 3 ms before and after N9 peak. As results, the network could distinguish each patient and normal subject with probability of 100% in both methods, but method using the secondary differential waveform was superior to another in the learning capability of the network.