Validating the Cognitron online cognitive assessment technology for large scale remote citizen science, clinical trials and individual assessment applications

Cognition is a vital part of the human experience and when it becomes impaired, it can have major impacts on quality of life. Therefore, the ability to sensitively measure cognitive functioning is an important tool in understanding how the human brain works. This thesis describes the development and deployment of a software package called “The Cognitron” that is designed to serve cognitive tasks and questionnaires in a wide variety of research and clinical settings, both online and in-person. Beyond a description of the software itself, three examples of its deployment in different research environments are reported. Firstly, cognitive and personality profiles of different psychiatric and neurological conditions are investigated in a large online remote citizen science project with over 800,000 self-selecting volunteers. Sophisticated and novel methods for cleaning such a dataset are described. The software shows good sensitivity to cognitive impairment, consistent with prior literature. Hierarchical clustering of mean profiles show good separation of psychiatric and neurological conditions. SVMs were trained on individual volunteer profiles to classify disease against healthy controls, with the psychiatric conditions showing highest classification accuracy. Secondly, the effects of SARS-CoV-2 on cognition are investigated in a highly controlled in-person prospective clinical challenge study in which 36 volunteers were intentionally infected with SARS-CoV-2 and were followed up periodically over a year. The results show a subtle but significant deficit in cognitive ability up to a year post inoculation. Thirdly its use in a traumatic brain injury clinic as a tool for assessing cognitive impairment is investigated. The results show high sensitivity of the software to cognitive impairment in traumatic brain injury. They also show that patients tested remotely do not perform statistically differently to those tested in-person. They also highlight heterogeneity in cognitive impairment in this clinical population, indicating the need for sensitive cognitive testing technology.