The ISC Cognitive Model: Automated and comprehensive interpretation of cognitive functioning

Clinicians aim to interpret a patient’s performance on a (set of) cognitive tests in terms of the underlying cognitive processes. For example, an interpretation of poor performance on a word recall test is only informative when it goes beyond the raw test score and provides insight regarding memory functioning. Digital testing offers the potential to automate the interpretation process. However, providing a standardized mapping between test scores and cognitive domains is not trivial. No cognitive test provides a perfect measure for a single cognitive domain. In fact, performance on most cognitive tests involves multiple cognitive domains.

Philips IntelliSpace Cognition (ISC) provides a comprehensive measurement of cognitive functioning that is driven both by neuropsychological theory and by data. A cognitive overview interprets a patient’s test performance in terms of established cognitive domains that go beyond the scores on individual tests. The cognitive domain scores are calculated using the ISC Cognitive Model, which is based on Structural Equation Modeling (SEM). The central idea of this SEM approach is that the scores from certain cognitive tests cluster together as they are caused by the same underlying cognitive domain. Even though the cognitive domains are not measured directly, the model estimates them mathematically by accounting for the relations between outcome measures. The model’s exact input measures, number of existing cognitive domains, and how these all relate are fully based on established neuropsychological literature and the expert opinion of several experienced neuropsychologists. Within the constraints of this theory-driven framework, the model searches for an optimal mathematical solution. Additionally, the model is able to automatically search for additional patterns in the data that in many cases are difficult to detect manually. For example, the ISC cognitive model found that a portion of Trail Making Test variance was related to the working memory domain, despite there being no direct link between TMT and working memory. This suggests that a part of Trail Making Test performance relies on Working memory. The model uses such patterns to break down test scores into distinct components that are associated with different cognitive processes.

Figure 1 presents a visualization of how the individual tests contribute to the cognitive domain scores. Conventional outcome measures of six cognitive tests provide the input for the model, resulting in the estimation of six cognitive domains: memory, working memory, processing speed, verbal processing, executive functioning, and visual spatial processing. Most tests contribute to at least two cognitive domain scores. This means that cognitive domains comprise multiple test scores, with greater contributions of test scores that form a purer measure of that cognitive domain. The Cognitive Model uses a normative data set consisting of 297 healthy adults. Sampling of the normative data was stratified by education level and race/ethnicity in line with the US census (CPS ASEC 2017). The Cognitive Model was found to provide an excellent fit to the data.”

Figure 1. Relative strength of the relationships between test scores and cognitive domains in the ISC Cognitive model. Darker shades represent stronger relationships