Variational autoencoder provides proof of concept that compressing CDT to extremely low-dimensional space retains its ability of distinguishing dementia

Sabyasachi Bandyopadhyay, Catherine Dion, David J. Libon, Catherine Price, Patrick Tighe, Parisa Rashidi

Research output: Contribution to journalArticlepeer-review

Abstract

The clock drawing test (CDT) is an inexpensive tool to screen for dementia. In this study, we examined if a variational autoencoder (VAE) with only two latent variables can capture and encode clock drawing anomalies from a large dataset of unannotated CDTs (n = 13,580) using self-supervised pre-training and use them to classify dementia CDTs (n = 18) from non-dementia CDTs (n = 20). The model was independently validated using a larger cohort consisting of 41 dementia and 50 non-dementia clocks. The classification model built with the parsimonious VAE latent space adequately classified dementia from non-dementia (0.78 area under receiver operating characteristics (AUROC) in the original test dataset and 0.77 AUROC in the secondary validation dataset). The VAE-identified atypical clock features were then reviewed by domain experts and compared with existing literature on clock drawing errors. This study shows that a very small number of latent variables are sufficient to encode important clock drawing anomalies that are predictive of dementia.

Original languageEnglish (US)
Article number7992
JournalScientific Reports
Volume12
Issue number1
DOIs
StatePublished - Dec 2022

All Science Journal Classification (ASJC) codes

  • General

Fingerprint

Dive into the research topics of 'Variational autoencoder provides proof of concept that compressing CDT to extremely low-dimensional space retains its ability of distinguishing dementia'. Together they form a unique fingerprint.

Cite this