ARTIFICIAL NEURAL NETWORKS FOR SCREENING PATIENTS NEEDING EMERGENCY CRANIAL COMPUTED-TOMOGRAPHY SCANS IN EMERGENCY DEPARTMENTS

Rationale and Objectives. We evaluated the potential for a neural netw ork to screen candidates for emergency cranial computed tomography (Cr ) scans in an emergency department setting. Methods. Data were collect ed from 1625 patients undergoing emergency cranial CT scanning in two different emergency departments (EDs). Singular value decomposition (S VD) was used to remap input data for network training. Data were rando mly divided into six subsets, and one was reserved as a test set to an alyze network performance. Five networks were then trained on data fro m the five remaining sets using fivefold cross-validation, Each traine d network was allowed an independent vote on need for CT scanning in e ach case from the test set. The majority vote was used as the final pr ediction. A similar analysis was done on data from each individual ED, Results are compared with prior statistical studies of the same data. Results. The network performed well when predicting clinical variable patterns that consistently produced negative CT scans and on patterns that were ambiguous in terms of the CT scan results. It performed poo rly, however, on patterns that consistently predicted positive scans, This last finding appears to have resulted from inadequate training ma terial. The taco populations from which data were taken were shown to be distinct, but a network trained on the combined data performed as w ell as the networks from the individual EDs in predicting patients req uiring CT scanning. Variables with the greatest contribution to the ne tworks' prediction were consistent with those in prior statistical stu dies. Conclusion. Although preliminary in nature, neural networks show promise as a screening device for selecting patients for emergent cra nial CT scanning.