A Deep Learning Approach to Predict Chronological Age
-
Published:2023-02-03
Issue:3
Volume:11
Page:448
-
ISSN:2227-9032
-
Container-title:Healthcare
-
language:en
-
Short-container-title:Healthcare
Author:
Lahza Husam1ORCID, Alsheikhy Ahmed A.2ORCID, Said Yahia2ORCID, Shawly Tawfeeq3ORCID
Affiliation:
1. Department of Information Technology, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah 21589, Saudi Arabia 2. Department of Electrical Engineering, College of Engineering, Northern Border University, Arar 91431, Saudi Arabia 3. Department of Electrical Engineering, Faculty of Engineering at Rabigh, King Abdulaziz University, Jeddah 21589, Saudi Arabia
Abstract
Recently, researchers have turned their focus to predicting the age of people since numerous applications depend on facial recognition approaches. In the medical field, Alzheimer’s disease mainly depends on patients’ ages. Multiple methods have been implemented and developed to predict age. However, these approaches lack accuracy because every image has unique features, such as shape, pose, and scale. In Saudi Arabia, Vision 2030, concerning the quality of life, is one of the twelve initiatives that were launched recently. The health sector has gained increasing attention as the government has introduced age-based policies to improve the health of its elderly residents. These residents are urgently advised to vaccinate against COVID-19 based on their age. In this paper, proposing a practical, consistent, and trustworthy method to predict age is presented. This method uses the color intensity of eyes and a Convolutional Neural Network (CNN) to predict age in real time based on the ensemble of CNN. A segmentation algorithm is engaged since the approach takes its input from a video stream or an image. This algorithm extracts data from one of the essential parts of the face: the eyes. This part is also informative. Several experiments have been conducted on MATLAB to verify and validate results and relative errors. A Kaggle website dataset is utilized for ages 4 to 59. This dataset includes over 270,000 images, and its size is roughly 2 GB. Consequently, the proposed approach produces ±8.69 years of Mean Square Error (MSE) for the predicted ages. Lastly, a comparative evaluation of relevant studies and the presented algorithm in terms of accuracy, MSE, and Mean Absolute Error (MAE) is also provided. This evaluation shows that the approach developed in the current study outperforms all considered performance metrics since its accuracy is 97.29%. This study found that the color intensity of eyes is highly effective in predicting age, given the high accuracy and acceptable MSE and MAE results. This indicates that it is helpful to utilize this methodology in real-life applications.
Funder
Institutional Fund Projects Ministry of Education and King Abdulaziz University, DSR, Jeddah, Saudi Arabia
Subject
Health Information Management,Health Informatics,Health Policy,Leadership and Management
Reference37 articles.
1. DNA Methylation Biomarkers-Based Human Age Prediction Using Machine Learning;Zaguia;Comput. Intell. Neurosci.,2022 2. Han, J., Kim, S.Y., Lee, J., and Lee, W.H. (2022). Brain Age Prediction: A Comparison between Machine Learning Models Using Brain Morphometric Data. Sensors, 22. 3. Hobday, H., Cole, J.H., Stanyard, R.A., Daws, R.E., Giampietro, V., O’Daly, O., Leech, R., and Váša, F. (2022). Tissue volume estimation and age prediction using rapid structural brain scans. bioRxiv. 4. Ramirez, J.G., Blazquez, M.A.F., and Rosa, J.J.G. (2022). Prediction of chronological age in healthy elderly subjects with machine learning from MRI brain segmentation and cortical parcellation. Brain Sci., 12. 5. Cao, Y., Berend, D., Tolmach, P., Amit, G., Levy, M., Liu, Y., Shabtai, A., and Elovici, Y. (2022, January 3–8). Fair and accurate age prediction using distribution aware data curation and augmentation. Proceedings of the 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA.
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
|
|