Login
Facebook YouTube Twitter Linkedin
Facebook YouTube Twitter Linkedin

Neural Network Classification of Longitudinal Cognitive Data for Prediction of Individual-Level

Clinical trials in Alzheimer's disease require enrichment of participants at the early stages of a disease, who are most likely to develop AD in the future. Recruiting such individuals remains a key challenge.

Our ability to predict whether an individual with mild cognitive impairment (MCI) will transition to Alzheimer’s disease (AD) or remain stable could be augmented by using automatic techniques that learn the complex temporal relationships between symptom presentation and future disease outcomes.

Methods

Participant with suspected MCI were recruited from a longitudinal study of neuropsychological function in community-residing adults. Participants were assessed on an extensive battery of CANTAB tests at baseline, ten months and 20 months as part of a comprehensive neuropsychological assessment.

The analysis sample consisted of 106 participants, 57% of whom were females with a mean age of 70 years and mean episodic memory scores (CANTAB PAL total errors adjusted) at baseline of 33 (SD 20). At baseline 25% were controls, 27% non-amnestic MCI, 48% amnestic MCI.

Results

The results of this research, presented at AAIC 2018, indicate that our algorithms showed excellent sensitivity for predicting MCI progression using PAL test scores. These findings also suggest that using automatic learning of complex temporal relationships in cognitive test scores could support specialists in making early diagnoses of Alzheimer’s disease, which could potentially enhance efficacy for the next wave of clinical trials.

Interested in learning more about how neural network classification of longitudinal cognitive data can be used to predict individual-level change? Download this poster.

 

Download here

Share this