Panoramix: detecting cognitive impairment through digital games

This article was published on behalf of the authors: Luis E. Anido-Rifón & Manuel J. Fernández Iglesias, Professors, atlanTTic – University of Vigo, Spain.

The atlanTTic research center of the University of Vigo proposed one of the pilot projects selected in the first Gatekeeper Open Call. The pilot aimed to exploit Gatekeeper’s Data Federation to manage the information captured by the Panoramix battery of digital games for the detection of cognitive impairment, as well as to integrate the games into the Gatekeeper marketplace.

Adressing the challenge of detecting cognitive decline

The detection and diagnosis of cognitive impairment is currently based on a set of standardised questionnaires (Clock Drawing Test, Isaacs Set-test, Pfeiffer Portable Mental State Questionnaire, Mini-Mental State Examination, …). There are other emerging techniques, such as positron emission tomography, which can detect characteristic plaques in brain tissue long before symptoms appear, and tests based on cerebrospinal fluid biomarkers. In any case, current standard practice, especially in primary care or outpatient clinics, is based on the administration of the tests above at signs of deterioration, typically detected by family members or caregivers.

The more traditional mechanisms used are time-consuming, expensive and have several limitations that must be taken into account to avoid biases in the results obtained. In general, personalised attention by professionals is required; many of the subjects evaluated perceive the sessions in which the tests are administered as intrusive and extraneous, and they often offer delayed detection, because the cognitive deficits are already evident when medical assistance is requested.

This delay clearly limits intervention and treatment options. In addition, most tests are not perceived as natural or close to the daily activities of the tested subject (i.e., they lack ecological validity). For example, neuropsychological assessments with pencil and paper are moderately correlated with daily activities and recurrent assessments with these tests lead to confounding of results due to practice effects. Finally, many cognitive assessment tools are not valid for people with comorbidities or low educational attainment.

In this context, Panoramix was born as a project to detect cognitive impairment overcoming the limitations of classical techniques. The development of Panoramix started in 2015 with the support of several R&D projects. The key milestone was the PhD Thesis of Sonia Valladares Rodriguez, then a young researcher at the atlanTTic research center. She conceptualised to whole Panoramix instrument and supervised the projects that supported its development. Her research in the field of machine learning allowed us to identify what information is relevant to detect the cognitive reserve of a person playing digital games.

Panoramix’ team set the goal of obtaining an ecological instrument, in the sense that it is accepted by the evaluated subjects and is not perceived as a hostile test, more related to daily activities, more robust against the effect of practice, and easier to administer. All this, guaranteeing a discrimination capacity of cognitive impairment and Alzheimer’s disease at least as good as that of the classic tests.

Panoramix implementation of the California Verbal Learning Test as a stroll through an European village (Source: atlanTTic – University of Vigo)

Panoramix was tested in the framework of several research and transfer projects by more than 130 elderly people in four countries with different degrees of cognitive impairment. As a consequence of the results obtained, Panoramix was selected by Gatekeeper’s first Open Call as one of the disruptive technologies to be included in the platform.

Panoramix shortens the time to detect Alzheimer’s disease

By eliminating pencil and paper questionnaires and bringing interaction with subjects into the digital world, we can present stimuli and capture reactions much more effectively, without the intervention of the examiner. In addition, we can collect at the same time other relevant information beyond the subject’s reactions to the challenges posed by the digital test, such as reaction time, fatigue, etc.

In addition, the analysis of the information obtained during the test and uploaded to Gatekeeper (i.e., reactions and responses of the subject) is performed automatically, which also contributes to significantly reduce screening time.

A lady playing with Panoramix to evaluate episodic memory during the Gatekeeper pilot (Source: atlanTTic – University of Vigo)

Shortening administration times is perhaps the most important feature from an economic and logistical point of view, as well as opening the door to other uses, such as, for example, carrying out screening campaigns at a reasonable cost or implement cognitive triage in outpatient clinics. However, as indicated, Panoramix benefits go much further.

The video game can uncover hidden signs of mental impairment within 40 minutes.

Throughout our research over the past six years, we improved the games to make them friendlier and more attractive to senior adults. In addition, by applying artificial intelligence techniques, and more specifically machine learning, we identified which are the most informative variables that allow us to detect the degree of cognitive impairment from a broad set of information captured from the subject’s interaction with the games. Moreover, from these most informative variables, we selected a reduced set that allows us to detect the degree of impairment with a quality at least as good as the classical tests.

In other words, our research goes far beyond the design of attractive games to maintain cognitive reserve, but includes an exhaustive and rigorous analysis of all the information obtained during the game to finally identify exactly what information is relevant as an indicator of cognitive impairment. This is perhaps the most relevant contribution.

Around a hundred older adults with a wide range of cognitive states, from people without cognitive impairment to people with Alzheimer’s disease, participated in this journey. In addition, professionals from the socio-health and medical fields collaborated with us, contributing their knowledge and experience to achieve the best possible tool.

Through trials with about a hundred subjects, we demonstrated that the discrimination power of our tool between healthy people, people diagnosable with mild cognitive impairment and people with Alzheimer’s is at least as effective as classical methods.

A lady playing with Panoramix to evaluate procedural memory during the Gatekeeper pilot (Source: atlanTTic – University of Vigo)

Gatekeeper provided us with a unique opportunity to conduct a first longitudinal pilot to investigate the cognitive evolution of a set of 30 subjects over time using Panoramix. The FHIR endpoint provided by Gatekeeper was used to collect and classify gaming data from those subjects over several months. Some of those subjects were performing supervised cognitive training tasks during the study period, while others chose not to. The analysis of the collected data allowed us to study the influence of such activities on the cognitive evolution of the subjects, as a function of their general cognitive state. In addition, the study paves the way for analysing the predictive power of Panoramix to estimate the future onset of cognitive impairment.