"Integrating theory-guided and data-driven approaches for measuring consciousness" Dr. Nao Tsuchiya
Professor Nao Tsuchiya, PhD
School of Psychological Sciences
Monash University, Australia
Title: Integrating theory-guided and data-driven approaches for measuring consciousness
Clinical assessment of consciousness is one of the most significant issues. Recent research suggests that some brain-damaged patients who are assessed as unconscious are actually conscious. The prospect of misdiagnosis of consciousness is terrifying. Misdiagnosis also happens when consciousness needs to be deliberately and reversibly suppressed by general anaesthesia. A failure of general anaesthesia, is a global clinical problem known to lead to numerous psychological problems, including post-traumatic stress disorder. Avoiding awareness with overdose of anaesthetics, however, can lead to various cognitive impairment.
Currently available objective assessment of consciousness is limited in accuracy or requires expensive equipment with major barriers to translation.
Previous approaches to consciousness detection have mainly taken one of two approaches: either data-driven or theory-guided. Data-driven approaches learn patterns in neural data that are informative for detecting consciousness by training machine-learning algorithms on large labelled datasets. Among the theories of consciousness, Integrated Information Theory (IIT) has emerged a highly promising candidate. IIT propose a theory-guided measure, Integrated Information (II). II can detect consciousness at high accuracy by simultaneously quantifying the degree of differentiation and integration in the brain. While promising, both data-driven and theory-guided approaches for consciousness detection remain limited.
In this talk, we will outline our recent theory-guided and data-driven approaches to develop new, optimized consciousness measures that will be robustly evaluated on an unprecedented breadth of high-quality neural data, recorded from the fly model system. We will overcome the subjective-choice problem in data-driven and theory-guided approaches with a comprehensive data analytic framework, which has never been applied to consciousness detection. We will integrate previously disconnected streams of research in consciousness detection to accelerate the translation of objective consciousness measures into clinical settings.