Our research

Personalized diagnostic testing is a promising venture in many academic fields. It is well established that it increases sensitivity, and there exist a multitude of incentives to develop such tools for wider use in medical fields particularly. The current report describes one such enterprise, namely Mindify, which aims to develop an online tool for online, global personalized testing for cognitive and memory impairment. This is done through a face recognition task, wherein participants are rated on their ability to correctly recall the names of celebrities. It is expected that this presents an opportunity to employ artificial intelligence methods to predict, on a personal level, changes in memory performance and thus potentially identify markers for early memory decline, when intervention is most effective. This project is currently in the first stages of testing, and must be optimized in order to guarantee its relevance for the wider population. The task was therefore to adjust the set of stimuli used so as to improve the accuracy to an acceptable level amongst a sample of young adults. The results showed

The Mindify Project

The Mindify project involves creating an online diagnostic test that would ideally be able to preemptively diagnose cognitive decline. Degenerative brain disorders and particularly memory disorders, such as Alzheimer’s Disease (AD), are of interest here. Diagnosing such disorders has proven to be troublesome, as the diagnostic process often only starts after severe symptoms appear, i.e when the disease has advanced too far for efficient countermeasures. Early detection is essential for patients, yet it is difficult to achieve this in practise. Despite the possibility of increasing diagnostic sensitivity to personal changes in cognitive capacity, tests are hardly ever made in a way that allows for personalization. Therefore, the idea of this project was to make a diagnostic test platform (Mindify), available online for public use, with a future aim to predictively analyse participants’ performance, which would enable both earlier detection and more personalized diagnosis.

One particular example of the tests available on the Mindify platform is the Mind Labs Memory Test, which aims to detect changes in memory capabilities across lifespan. This online questionnaire uses images of celebrities, which participants must attempt to name. Importantly, the test is adaptive, so that it presents a series of images predicted to result in about 70-80% recall (naming correctly). No future meaningful analysis would be possible if the majority of assumably healthy participants obtained low scores, because there simply wouldn’t be enough variance between them and individuals who are suffering from cognitive decline. Therefore, optimizing the test performance early on is a vital step in order to ensure the applicability of the tool in a wider population.

Theoretical Framework

Aging is often associated with some cognitive decline. Difficulties arise in many areas, but of particular interest here, is the decline in memory capacities. Mild memory loss often represents nothing but the normal process of aging, but in some cases, forgetting might be indicative of more serious memory disorders. Memory loss is one of the earliest symptoms of AD, but it is initially virtually indistinguishable from normal aging (Morris, et al., 2001), thus herein lies the technical difficulty in targeting the disease. However, knowing that recently formed memories tend to deteriorate before more remote memories are affected (Orlovsky, et al., 2017), early detection might be possible by leveraging this fact. Artificial Intelligence (AI) methods could successfully evaluate participants, not on the general norms, but rather on their personal capacity to correctly recall the name of a person they’re likely to know or have known.

Therefore, of particular interest is a Harvard study (Germine, et al., 2012), where images of famous faces were used in order to collect data relating to changes in cognitive capacities. Although this isn’t necessarily directly related to online diagnostics, it is still a comparison between web and laboratory experiments. Therefore, a very similar approach was taken in this project, and the MLMT also uses images of celebrities to collect data on recall capacities, which can later be related to demographic information. The dataset itself is nevertheless different, as the Harvard dataset excluded all information, except the face (so no hair or background, contrary to the dataset used here).

References

Chawla, Nitesh & Davis, Darcy. (2013). Bringing Big Data to Personalized Healthcare: A Patient-Centered Framework. Journal of general internal medicine. 28(3). 10.1007/s11606-013-2455-8.

Germine, L., Nakayama, K., Duchaine, B., Chabris, C., Chatterjee, G., & Wilmer, J. (2012). Is the web as good as the lab? Comparable performance from web and lab in cognitive/perceptual experiments. Psychonomic Bulletin & Review, 19(5): 847-857.

Morris, J. C., Storandt, M., Miller, J. P., McKeel, D. W., Price, J. L., Rubin, E. H., & Berg, L. (2001). Mild cognitive impairment represents early-stage Alzheimer disease. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/11255443

Orlovsky, I., Huijbers, W., Hanseeuw, B. J., Mormino, E. C., Hedden, T., Buckley, R. F., … Papp, K. V. (2017). The relationship between recall of recently versus remotely encoded famous faces and amyloidosis in clinically normal older adults. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5956796/

Yu, A. Z., Ronen, S., Hu, K., Lu, T., & Hidalgo, C. (2016, January 4). Pantheon 1.0, A Manually Verified Dataset of Globally Famous Biographies. Retrieved from https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/28201