Search

Additional file 1: Table S1. Demographics of the participants who were non-demented (NDE, i.e., individuals with normal cognition (NC) and mild cognitive impairment (MCI)) at the time of the voice recordings. ApoE data was unavailable for one Gen 1 participant, thirteen Gen 2 participants, and one New Offspring Cohort (NOS) participant; MMSE data was not collected for all Gen 3, OmniGen 2, and NOS participants.

Additional file 5: Table S2. For the neuropsychological tests that have too few samples, the average salient administered fraction (SAF) and standard deviation for true positive (SAF[+]) and true negative (SAF[-]) cases are listed in descending order based on the SAF[+] value. SAF[+] is calculated by summing up the time spent in a given neuropsychological test that intersects with a segment of time that is DE[+] salient and dividing by the total time spent in a given neuropsychological test. SAF[-] is calculated by summing up the time spent in a given neuropsychological test that intersects with a segment of time that is not DE[+] salient and dividing by the total time spent in a given neuropsychological test. The number of samples for SAF[+] and SAF[-] indicate the number of true positive and true negative recordings that contain each neuropsychological test.

BACKGROUND: Widespread dementia detection could increase clinical trial candidates and enable appropriate interventions. Since the Clock Drawing Test (CDT) can be potentially used for diagnosing dementia-related disorders, it can be leveraged to develop a computer-aided screening tool. OBJECTIVE: To evaluate if a machine learning model that uses images from the CDT can predict mild cognitive impairment or dementia. METHODS: Images of an analog clock drawn by 3,263 cognitively intact and 160 impaired subjects were collected during in-person dementia evaluations by the Framingham Heart Study. We processed the CDT images, participant’s age and education level using a deep learning algorithm to predict dementia status. RESULTS: When only the CDT images were used, the deep learning model predicted dementia status with an area under the receiver operating characteristic curve (AUC) of 81.3% ± 4.3%. A composite logistic regression model using age, level of education, and the predictions from the CDT-only model, yielded an average AUC and average F1 score of 91.9% ± 1.1% and 94.6% ± 0.4%, respectively. CONCLUSION: Our modeling framework establishes a proof-of-principle that deep learning can be applied on images derived from the CDT to predict dementia status. When fully validated, this approach can offer a cost-effective and easily deployable mechanism for detecting cognitive impairment.