Technology-assisted approaches are mandatory to keep up with the diagnostic requirements in terms of both quality (objective, reliable and reproducible observations) and quantity (ever increasing number of patients exasperated by the prospect of a relative decrease of professional resources). Studying interaction patterns to initiate or adjust deleterious or favor positive interaction patterns in settings where several patients and other people interact with someone having dementia and integrate them using objectively measured physiological stress and behavioural parameters might greatly help increase diagnostic accuracy and, consequently, improve treatment for and help given to the individual patient.
To do so, real-time observations and measures are required. This is very different from current standards applied in clinical settings where only few-point observations made by nursing or medical staff are available, observations that critically depend on adequate staffing and staff competence. Most technology-assisted measures in clinical medicine do not take into account the typically human meso-temporal context, defined here as the time frame extending from a few minutes to a few days within the habitual context of life of an individual person or patient.
From Large Data to Actionable Extracts
Our overall approach is based on the assumption that indexing of physiological, behavioral and topological (PBT) “meso-patterns” provides objective and quantitative measurements of individual and disease-related features as well as response to treatments. The concept itself stems from and relies on mathematical frameworks yielding macroscopic characterizations of physical systems based on their microscopic properties. Such an approach has been used many times in various scientific disciplines. In our approach, the molecular level is represented by tiny changes in PBT readings that we map onto a Hurst index or Hausdorff-Besicovitch dimension that can be thought of as a macroscopic system parameter.
Artificial Intelligence and Predictability
At least some of the indices considered such as, e.g., heart and respiratory rates are unique for each individual. We assume that this observation would also apply to the other proposed indices given their stochastic nature. This means that a collection of complex indices can provide aggregated characterization of factorized groups of patients. This leads to “fingerprinting”, i.e. characterization of an individual. Consequently, the correlation between factorized groups and indexing of their responses to treatment provide the basis for personalized medicine. Ultimately, we are able to index individual behavior in humans, using a behavioral vector to brain connectivity.