ObjectiveTo identify which combination of motor evoked potentials (MEPs) and somatosensory evoked potentials (SEPs) is most reliable for postoperative motor deterioration during spinal cord tumor surgery, according to anatomical and pathologic type.MethodsMEPs and SEPs were monitored in patients who underwent spinal cord tumor surgery between November 2012 and August 2016. Muscle strength was examined in all patients before surgery, within 48 hours postoperatively and 4 weeks later. We analyzed sensitivity, specificity, positive and negative predictive values of each significant change in SEPs and MEPs.ResultsThe overall sensitivity and specificity of SEPs or MEPs were 100% and 61.3%, respectively. The intraoperative MEP monitoring alone showed both higher sensitivity (67.9%) and specificity (83.2%) than SEP monitoring alone for postoperative motor deterioration. Two patients with persistent motor deterioration had significant changes only in SEPs. There are no significant differences in reliabilities between anatomical types, except with hemangioma, where SEPs were more specific than MEPs for postoperative motor deterioration. Both overall positive and negative predictive values of MEPs were higher than the predictive values of SEPs. However, the positive predictive value was higher by the dual monitoring of MEPs and SEPs, compared to MEPs alone.ConclusionFor spinal cord tumor surgery, combined MEP and SEP monitoring showed the highest sensitivity for the postoperative motor deterioration. Although MEPs are more specific than SEPs in most types of spinal cord tumor surgery, SEPs should still be monitored, especially in hemangioma surgery.

Various information tools appeared to help users locate and retrieve information resources over the Internet. The amount of information resources is increasing dramatically, and it is necessary to use indexing services to deal with the scalability problems. Existing indexing systems collect index data periodically from information servers only. Thus the users of indexing systems lack a common framework where they can share their heuristic information on available resources. This paper introduces collaborative indexing for gathering index data from users. The indexing system is seamlessly integrated with existing information tools. In our model, a group of users with a common view shares feedback of resource discovery. One user's feedback about searches is used by future users to help locate relevant resources. We implemented a prototype system where agents share index data by weighted association. Each agent de nes its domain by community and topic the agent serves. Agents gather users' feedback and use it to change weight values of associations. By doing this, users are relieved from a large volume of irrelevant resources in a global search space, and get more closely related matches to the search request.