Research Article
Detection procedures for Patients of Alzheimer’s Disease using Waveform Features of Pupil Light Reflex in response to Chromatic Stimuli
@ARTICLE{10.4108/eai.17-12-2020.167656, author={Wioletta Nowak and Minoru Nakayama and Tomasz Kręcicki and Andrzej Hachoł}, title={Detection procedures for Patients of Alzheimer’s Disease using Waveform Features of Pupil Light Reflex in response to Chromatic Stimuli}, journal={EAI Endorsed Transactions on Pervasive Health and Technology}, volume={6}, number={24}, publisher={EAI}, journal_a={PHAT}, year={2020}, month={12}, keywords={Pupil, Pupil Light Reflex, Alzheimer’s Disease, Sparse estimation, logistic regression}, doi={10.4108/eai.17-12-2020.167656} }
- Wioletta Nowak
Minoru Nakayama
Tomasz Kręcicki
Andrzej Hachoł
Year: 2020
Detection procedures for Patients of Alzheimer’s Disease using Waveform Features of Pupil Light Reflex in response to Chromatic Stimuli
PHAT
EAI
DOI: 10.4108/eai.17-12-2020.167656
Abstract
INTRODUCTION: Various studies conducted to predict Alzheimer’s disease (AD) indicate that some pupillight reflex (PLR) features may contain symptoms of the disease. An effective procedure that can predict thedisease using PLRs is needed.
OBJECTIVES: Two analytic approaches were examined in order to estimate the possibility of identifying Alzheimer’s patients using features of PLR waveforms from chromatic stimuli. In particular, an index of the probability of being an AD patient is introduced, and the features which contributed to PLRs the most were extracted.
METHOD: PLRs for three colours of light pulses (red: 635nm, blue: 470nm, white: CIE x=0.28, y=0.31) at twolevels of intensity (10 and 100 cd/m2) were observed at 60Hz for 10s. Pulses consisted of pre-stimulus (2s), light pulse (1s) and restoration phases (7s). 15 features were extracted from each PLR waveform, such as pupil constriction velocity, pupil response delay, etc. Seven AD patients (age:42-84, mean=68.1) and 12 similar-aged control subjects (age:62-89, mean=72.1).
RESULTS: The first approach was based on factor scores of features of PLRs. Two factor scores were extracted from the 15 features across all measurement conditions, and logistic functions were introduced in order to calculate the probability of identifying AD patients. Function parameters were estimated using a Bayesian technique, such as the Markov chain Monte Carlo method (MCMC). In consideration of the number of participants and biased data distributions, the second approach was based on the sparse modelling technique. Least absolute shrinkage and selection operator (LASSO) was applied to sets of PLR features from each light stimulus, together with the ages of subjects, and optimised result sets were obtained. Prediction performance was higher than with the previous procedure.
CONCLUSION: The use of PLRs features from chromatic stimuli for identifying AD was developed and evaluated.
Copyright © 2020 W. Nowak et al., licensed to EAI. This is an open access article distributed under the terms of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/), which permits unlimited use, distribution and reproduction in any medium so long as the original work is properly cited.