System health management and space medicine predictive diagnostics. Common concepts and approaches

Mission design for future human exploration spaceflight (moon, asteroids and Mars), with their inherent risks and communications delays, requires a shift in space medicine from a telemedicine paradigm to that of medical autonomy. With the absence of real-time medical ground support, clinical decision support technologies for health monitoring and diagnostics may be required to assist the onboard Crew Medical Officer (CMO). This paper introduces a predictive diagnostics concept for space medicine based on an approach which could be common for both biological and technical systems. In discussing the similarities and differences between the two domains, the paper explores potential solutions applicable for health risk assessment and management. Prognostics and Health Management (PHM) and associated challenges can be represented in different terms. The engineering discipline focuses on the fundamental principles of system failures in attempt to predict when they might fail, and links the principles to system life cycle management. This is referred to as Prognostics and Health Management (PHM). Since PHM is a maturing discipline the scientific method that underlines all scientific disciplines has not made its way into PHM research [1]. It must be noted also that there is no general agreement in the research community with respect to the PHM terminology (for example, component aging). [1, 2] Like PHM, predictive diagnostics in conventional medicine is in its infant state due to the lack of scientific foundation on disease prediction. Despite of the infancy stage both of the disciplines have much in common. For both, it is important to know what factors determine the occurrence of the failure (disease). In predictive diagnostics, identification of negative trends with particular failure precursors followed by predicting the future health condition (if no preventive measures are taken) is a challenging task. While prognostics is an engineering discipline focused on predict- ng the time at which a system or a component will no longer perform its intended function [3], predictive diagnostics is built on the powerful foundation of predictive analytics. Unlike prognostics, predictive analytics in space medicine would need to take into account the essential element - the uniqueness of each crew member. It requires the development of a “set of fingerprints” for each individual that would include information on medical history with the individual health baseline, genetic pre-disposition, medical events, operational context, etc. Whereas predictive analytics indicates what is going to fail, predictive diagnostics also indicates both root and contributing causes of, or factors leading to the potential failure, as well as a priority of the impending failure. However, like PHM, the predictive analytics is based on health maintenance, eliminating the “surprise” factor. The paper introduces a concept for space medicine predictive diagnostics based on real-time health monitoring. To leverage the strength from both data-driven and model-based approaches, a hybrid approach was taken. The concept includes a module-based architecture consisting of a space medicine domain knowledgebase, a customizable inference engine supported by a number of “plug-in-play” voting inference sub-engines, and a web-based interface.