Training Response Artificial Intelligence Network (TRAIN)

Using Artificial Intelligence to Predict Cognitive Training Response in Amnestic Mild Cognitive Impairment

This study will train a machine learning tool to predict response to a cognitive training intervention using baseline brain MRI sequences from older adults with amnestic mild cognitive impairment.

This randomized controlled clinical trial proposes to test a well-supported Cognitive Training intervention approach for enhancing cognition in older adults who meet criteria for amnestic mild cognitive impairment (aMCI). The trial will recruit participants recently enrolled in a larger multi-site clinical trial (NCT04171323) at the University of Florida site based on meeting inclusion criteria and willingness to participate in the additional procedures (baseline MRI) required. The intervention will consist of a take-home, iPad-based, adaptive cognitive training (Posit Science Brain HQ Suite) program which participants will be trained to self-administer 2-3 days/week for 60-minutes per session over the course of 16 weeks with a goal of completing 40 training sessions total. Participants randomized to one of four arms of cognitive training or an active control condition. All cognitive training arms will include training on a measure of rapid decision making (Double Decision, similar to the Useful Field of View [UFOV] task) in addition to a combination of other subtests to determine which combination best facilitates cognitive gains. Outcomes: The investigators plan to test the hypotheses that 1) Compared to the active control condition, CT will result in greater than 1.0SD cognitive improvement on several trained tasks (proximal transfer) and untrained cognitive skills (near transfer) for 20% of the sample, while another 20% will continue to experience cognitive decline equivalent to 1.0SD or greater despite adequate adherence to the intervention, and 2) baseline factors related to cognitive function, brain structure, and resting-state brain function can ultimately predict which individuals will show improvement after the 16-week intervention. These baseline factors will be measured by a well-established neuropsychological battery, the useful field of view (UFOV) task, a detailed set of MRI sequences, and their associations with the outcomes will be determined by a type of machine learning known as multi-modal support vector machine (SVM). This type of analysis will allow the integration of a myriad of data points and MRI features to combine into a highly accurate predictive model for cognitive functioning improvement after intervention. The primary outcome measure that will be predicted by the SVM is the UFOV composite score change, which is a combination of post-minus-pre intervention change across UFOV processing speed, selective attention, and divided attention scores. The secondary outcome measure that will be predicted by the SVM is the Posit Composite score, which is a fixed difficulty assessment of proximal transfer effects of the trained tasks. Assessments will occur at baseline (month 1) and at the end of the 16-week intervention (Month 4). Brain imaging will only be performed at baseline. The primary aim is to further establish the efficacy of this well-supported cognitive training paradigm in a population of patients with amnestic Mild Cognitive Impairment, and the secondary aim is to distinguish intervention responders from non-responders using baseline cognitive, functional-neural, and demographic data alone.

Participants will complete computerized cognitive training via Posit Science Brain HQ. The duration is 60 min/day; the frequency is two to three days/wk, for 16 weeks with the goal of completing 40 sessions.

Participants will complete cognitively-stimulating computer activities. The duration is 60 min/day; the frequency is two to three days/wk, for 16 weeks with the goal of completing 40 sessions.

The intervention group will receive cognitive training that provides specific training tasks directed at essential cognitive domains tied to the project's aims and hypotheses (processing speed and working memory) that correspond with the neurocognitive and functional neuroimaging measures to be studied.

Participants will complete cognitively-stimulating computer activities. The duration is 60 min/day; the frequency is two to three days/wk, for 16 weeks with the goal of completing 40 sessions.

The Useful Field of View (UFOV) composite change score is calculated by subtracting the post-intervention score (0-500) from the pre-intervention score (0-500), where a more negative change score is indicative of stronger improvement after intervention (range +500 to -500).

Inclusion Criteria: Evidence of changed cognition reported by a close informant as obtained by an interview with a licensed clinical psychologist (PI, Gullett) Cognitive performance on one or more standardized neuropsychological measures of verbal or non-verbal memory > 1.0 standard deviation or more below the normative mean for age and education No evidence of reported dependence in instrumental functional abilities (IADLs) despite reported cognitive difficulties No evidence of dementia based on cognitive screening (Montreal Cognitive Assessment Score within normal limits for age, education, and sex using the NACC Uniform Data Set (UDS) norms. Global clinical dementia rating scale (CDR) score must be equal to 0.5 Willingness to be randomized into either the educational training (control) or intervention (cognitive training) group Able and willing to devote 12 weeks to the intervention with additional time for pre- and post-testing Able and willing to perform cognitive and emotional measures both on paper and on a tablet computer In line with the recommendations of the SCD task force, an informant must be available. This is to a) obtain information about the participant's complaints and cognitive difficulties on interview, and b) to ascertain validity of self-reported IADL function in regard to instrumental functional activities. Exclusion Criteria: Sensory loss (vision, hearing) or motor deficits that would preclude participation in the experimental tasks or neuropsychological assessment English as a second language Inability to undergo brain imaging due to claustrophobia or implants such as pacemakers, heart valves, brain aneurysm clips, orthodontics, non-removable body jewelry, or shrapnel containing ferromagnetic metal Previous major strokes or other known significant brain abnormalities or diseases affecting the brain and/or cognition (e.g., Parkinson disease, multiple sclerosis, seizure disorder, brain surgery, moderate TBI, REM Behavior Sleep Disorder, untreated sleep apnea, etc.) Unstable and uncontrolled medical conditions (metastatic cancer, HIV, moderate-severe kidney disease, uncontrolled diabetes, uncontrolled hypertension, severe cardiac disease, etc.). No current cancer diagnosis Current or past history of major psychiatric disturbance including schizophrenia, or active psychosis, bipolar disorder, current major depressive episode, current alcohol or substance abuse or history thereof within the past six months. The investigators are not excluding individuals who are taking antidepressant or anti-anxiety medications, however, use of these medications will be recorded and data analyzed in post-hoc analyses Use of antipsychotics, sedatives, or other medications with significant anticholinergic properties (due to potential influence on memory) Use of photo-sensitive medications such as steroids or retin-A within 15 days of the study intervention. Previous participation in a cognitive training study within the last 6 months or current involvement in another study involving cognitive, physical or other intervention at the time of participation.