Electrodiagnostic Nerve Tests: Understanding Healthy Peripheral Nerves

Abstract

The Nerve Excitability Test (NET) is an electrodiagnostic test capable of non-invasive characterization of peripheral nerves in humans. It has utility in differentiating between healthy controls and subjects with peripheral nerve disorders. Full realization of the diagnostic potential of NET requires a substantial database of normative values. This thesis describes the process of combining NET data from multiple centers around the world (Canada, n=120, 57 male, ages 18-70; Japan n=85, 50 male, ages 19-86; Portugal n=42, 14 male, ages 22-84) to create the first international normative NET dataset for the human median nerve. Since NET data often has missing values, we compared a number of approaches for filling missing values. An iterating cascading autoencoder performed the best. A simpler method, iterating linear regression, has similar performance while executing faster. The iterating cascading autoencoder was used to fill the missing values in the normative dataset. Data collected from multiple locations can suffer from “batch effects”: site-specific technical differences that reduce the homogeneity of the data. We developed a novel method for the detection of site-specific differences and found the homogeneity of the data from the three countries to be 95%, suggesting it is appropriate to combine the data into a single dataset. Comparison of the means of Canadian and Japanese NET data suggested that the remaining heterogeneity is due to technical differences in the stimulus-response curve, and that these differences have little or no impact on biological measures of nerve health. After establishing the normative dataset, we created a website which can be used as a clinical decision support system: NerveNorms.Bellstone.ca. The website also presents a nerve health score, interpretable as a p-value, to provide a quick and intuitive measure of the health of individual NET results relative to the normative dataset.