Scaffolding in AI-Empowered Cognitive Training Improves Outcomes in Cognitive Impairment

The general cognitive ability was assessed by the Cognitive Index, calculated from task performance scores. The CCT platforms recorded the Cognitive Index in real time on a daily basis. Weekly Cognitive Index values were computed as the mean of the daily scores recorded over each 7-day period. The change in Cognitive Index from baseline to week 8 is reported.

Seven cognitive subdomains include perception, attention, memory, language, executive function, thinking, and emotion. Domain-specific cognitive ability was calculated by averaging the scores of all tasks targeting the same cognitive domain. The change from baseline to week 8 is reported.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the Cognitive Index. Training dose was defined as the cumulative number of completed training tasks up to each training day. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in the Cognitive Index with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the Cognitive Index. Training dose was defined as the cumulative number of completed training tasks up to each training day. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which the Cognitive Index improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the perception subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in perception with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the perception subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the perception subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the attention subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in attention with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the attention subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the attention subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the memory subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in memory with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the memory subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the memory subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the language subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in language with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the language subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the language subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the executive function subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in executive function with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the executive function subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the executive function subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the thinking subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in thinking with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the thinking subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the thinking subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the emotion subdomain index. The outcome measure was the model-estimated asymptote parameter, representing the maximum attainable improvement in emotion with increasing training dose. Higher values indicate greater maximum attainable improvement.

A dose-response function was fitted to model the relationship between cumulative training dose and change from baseline in the emotion subdomain index. The outcome measure was the model-estimated learning-rate parameter, representing the rate at which performance in the emotion subdomain improved with increasing training dose. Higher values indicate a faster rate of improvement.