Pervasive Computing Technologies for Ambulatory Cognitive Assessment

Affective disorders cause mood disturbance in individuals and the main types are Bipolar Disorder (BD) and Major Depressive Disorder (MDD). Cognitive impairments in patients with affective disorders may indicate remission of symptoms, which can adversely affect quality of their daily tasks. Neuropsychological tests are the standard tools that are administered to the patients when they come for a follow-up visit.

Such tests assess various cognitive domains including memory, attention, processing speed, and executive functions. The follow-up visits are scheduled on a need basis and are administered in a silent room at a clinic by a trained staff. However, clinics have faced lack of resources as the number of patients with affective disorders is increasing.

Moreover, human cognitive functioning changes from time to time within a given day. Taken together, frequent assessments are essential to monitor patients’ cognitive functioning over time for timely diagnosis and early treatments. Two pervasive computing technologies were designed, implemented, and evaluated to 1) deliver a patient-administered cognitive assessment tool and 2) assess cognitive functioning in individuals’ free-living conditions.

Internet-based Cognitive Assessment Tool (ICAT) is a Web-based tool that automatically calculates cognitive test scores regarding verbal memory, working memory, and psychomotor speed. In particular, ICAT utilises speech recognition technology in verbal memory tasks. Hence, patients can take the tests at home without receiving any assistance from a clinician.

Ubiquitous Cognitive Assessment Tool (UbiCAT) is a wearable computing technology to 1) assess individuals’ attention, working memory, and executive functions over time using three smartwatch-based apps and 2) collect multivariate sensor data including activity and sleep features for digital phenotyping.

Three studies were conducted with ICAT to evaluate usability, feasibility, and concurrent validity of the test scores. The findings of these studies demonstrated high usability and significant validity of the test scores when compared with goldstandard neuropsychological tests. Three studies were also conducted with UbiCAT to investigate usability, validity, and feasibility of this tool in individuals’ free-living conditions.

As such, high usability of the UbiCAT and a strong correlation coefficient between UbiCAT and standard computerised cognitive tests were obtained. In addition, concurrent validity of the UbiCAT cognitive test scores was demonstrated when compared with neuropsychological tests in a population of healthy controls and patients with BD.

Our findings also proved feasibility of UbiCAT in accordance to the participants’ Global Positioning System (GPS) data, which showed that cognitive test performance measures were statistically the same in indoor and outdoor places. Supervised learning methods were applied on a dataset including one-week daily observations of cognitive, behavioural, and physiological features of controls and patients with BD for digital phenotyping.

As such, Extreme Gradient Boosting (XGBoost) model gave the highest performance and a set of digital phenotypes were derived from this model, showing that individuals’ time in bed, daily step counts, number of daily missed counts in the cognitive test sessions, and average of daily executive functioning are the most important features in determining their mental health diagnosis.

Overall, ICAT and UbiCAT are two pervasive computing technologies designed and implemented to provide ambulatory cognitive assessments. The findings of the studies conducted in this thesis demonstrated usability and feasibility of these tools as well as significant validity of their cognitive test scores.