Facial Recognition for Disease Diagnosis Using a Deep Learning Convolutional Neural Network: A systematic Review and Meta-Analysis

Special Article: Primary Healthcare

J Fam Med. 2024; 11(1): 1345.

Facial Recognition for Disease Diagnosis Using a Deep Learning Convolutional Neural Network: A systematic Review and Meta-Analysis

Xinru Kong1; Ziyue Wang2; Jie Sun3; Xianghua Qi4; Xiao Ding5*; Qianhui Qiu6*

1Shandong University of Traditional Chinese Medicine, Jinan City, Shandong Province, & Department of Otolaryngology, Guangdong Provincial People’s Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou City, Guangdong Province, China

2Shandong University of Traditional Chinese Medicine, Jinan City, Shandong Province, China

3Rizhao Central Hospital, Rizhao City, Shandong Province, China

4Affiliated Hospital of Shandong University of Traditional Chinese Medicine Jinan City, Shandong Province, China

5Affiliated Hospital of Shandong University of Traditional Chinese Medicine, Jinan City, Shandong Province, China

6Department of Otolaryngology, Guangdong Provincial People’s Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou City, Guangdong Province, China

*Corresponding author: Qianhui Qiu Department of Otolaryngology, Guangdong Provincial People’s Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou City, Guangdong Province, 510000, China; Xiao Ding, Shandong University of Traditional Chinese Medicine, Jinan City, Shandong Province, 250355, China. Tel: +86 18754418523 Email: qiuqianhui@hotmail.com; dxtcm2012@126.com

Received: December 06, 2023 Accepted: January 08, 2024 Published: January 15, 2024

Abstract,

Objective: This study aimed to systematically review the literature on facial recognition technology based on deep learning networks in disease diagnosis over the past ten years to identify the objective basis of this application.

Methods: This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for literature search and retrieved relevant literature from multiple databases, including PubMed, on November 13, 2023. The search keywords included deep learning convolutional neural networks, facial recognition, and disease recognition. 208 articles on facial recognition technology based on deep learning networks in disease diagnosis over the past ten years were screened, and 22 articles were selected for analysis. The meta-analysis was conducted using Stata 14.0 software.

Results: The study collected 22 articles with a total sample size of 57,539 cases, of which 43,301 were samples with various diseases. The meta-analysis results indicated that the accuracy of deep learning in facial recognition for disease diagnosis was 91.0% [95% CI (87.0%, 95.0%)].

Conclusion: The study results suggested that facial recognition technology based on deep learning networks has high accuracy in disease diagnosis, providing a reference for further development and application of this technology.

Keywords: Disease identification; Face recognition; Review; Convolutional neural network

Abbreviations: FCN: Fully Connected Convolutional Neural Network; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses; CNN: Convolutional Neural Networks; AHRQ: Agency for Healthcare Research and Quality; AI: Artificial Intelligence; ML: Machine Learning; DL: Deep Learning; AS: Angelman Syndrome; UPD: Paternal Uniparental Disomy; CdLS: Cornelia de Lange Syndrome; TS: Turner’s Syndrome; WBS: Williams-Beuren Syndrome; VGG: network: Visual Geometry Group Network; GSs: Genetic Syndromes; MTCNN: Multitask Cascaded Convolutional Neural Network; KNN: K-Nearest Neighbors; LDM: London Medical Database; NGP: Next Generation Phenotyping; SVM: Support Vector Machines; LM: Language Models; RT: Random Forests; EM: Expectation Maximization; SLR: Systematic Literature Review

Introduction,

Thousands of years ago, the traditional Chinese medicine book 'Huangdi Neijing' recorded that 'Qi and blood in the three hundred and sixty-five veins of the twelve meridians flow into the face and inject into the orifices (facial seven orifices) [1]. In China, many traditional Chinese medicine doctors can understand the patient’s condition by observing the patient’s face, which is called facial diagnosis. Diseases not only cause abnormalities in human internal structures and physiological functions but also may lead to changes in appearance and deformity, including changes in limbs, trunk, and face [2]. Doctors can make preliminary judgments on specific diseases by identifying the facial features of patients, which play a prompt role in the subsequent diagnosis and treatment of diseases. Hereditary diseases, such as Down syndrome [3], Turner syndrome [4], Nunan syndrome [5] and Williams-Beuren syndrome [6], Angelman syndrome [7], and Corneliade Lange syndrome [8], are the main causes of facial change or deformity. Meanwhile, there are some non-genetic diseases such as acromegaly [9], autism [10], and Alzheimer's disease [11]. Now, more new research focuses on cancer and the new coronavirus. However, the symptoms and facial features of some patients in the early stage of the disease are not prominent, and medical personnel lack experience in identifying the appearance of complex diseases and many other issues [12,13]. Therefore, before the development of convolutional neural-based face recognition technology, the facial features of the disease could not play an important role in the diagnosis of the disease.

In areas where medical resources are scarce, physical examination (especially large-scale equipment inspection) is a luxury, which leads to delays in disease treatment. Even in big cities, due to the difference in personal medical levels and the long queuing time of hospitalization, the treatment of diseases is often delayed [14]. Computer-aided facial diagnosis helps to quickly conduct noninvasive screening of diseases, which provides a reference for doctors to diagnose diseases timely and effectively [15].

Nowadays, face recognition technology has been preliminarily applied to the auxiliary diagnosis of the above diseases [16]. By analyzing the visual features of the face, face recognition technology identifies the identity for human-computer interaction, tracking and monitoring, security monitoring, and identity recognition. In the last ten years, this technology has gradually entered the field of medical diagnosis [17,18].

In this field, face recognition software determines the patient’s disease type by extracting the measurement data of specific features of the patient’s face, analyzing the patient’s face pattern, and comparing it with the disease database. Face recognition technology can identify diseases in time and alleviate the problem of insufficient medical resources. Now, for some diseases, face recognition technology even achieves higher diagnostic accuracy than doctors. Therefore, face recognition can contribute to the early screening of diseases and improve diagnostic accuracy.

This paper aims to provide a comprehensive overview of the current research on the application of face recognition technology to disease diagnosis. Previous surveys were extended by providing a detailed analysis of the history of face recognition and its evolution in the field of medical diagnosis [19]. Also, this study highlighted the potential of face recognition technology in overcoming the challenges of early disease screening and improving diagnostic accuracy. The rest of this paper is organized as follows: Section 2 introduces the meta-analysis design, data collection, and analysis methods. Section 3 provides a summary of the main findings for each study, including sample size, effect size, confidence interval, heterogeneity index, and an overall evaluation of the meta-analysis results for facial recognition in disease diagnosis. The discussion in Section 4 explains the importance of the meta-analysis results and suggests future research directions. Finally, Section 5 emphasizes the importance and contribution of facial recognition in disease diagnosis while acknowledging its limitations.

Material and Methods,

In the past ten years, the use of Convolutional Neural Networks (CNNs) for disease recognition has received significant attention from the scientific community. Therefore, this study reviewed publications from the past ten years (2013-2023) because there were fewer studies in this field ten years ago. Ethical approval to conduct the study was obtained from the Ethics Committee of Affiliated Hospital of Shandong University of Traditional Chinese Medicine (2021-074-YK).

Literature Search Strategy,

This systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). A literature search was conducted on December November 13, 2023, and the PubMed library, EBSCO Industries, Elsevier Scopus, Web of Science, and Springlink databases were utilized in this study. These databases were chosen because they provide the most high-impact scientific conference records and journals and cover the field of disease research and machine learning using CNNs. A broad search string was used to avoid missing any potentially exciting research. The search keywords were ("disease" or "disease recognition") and ("convolutional neural network" or "deep learning") and ("face recognition" or "facial diagnosis"). Additionally, other searches were conducted by reviewing the reference lists of all included papers. This comprehensive scope involved a large number of papers, and the most relevant ones were selected using the selection criteria outlined in Table 1. The studies included were cross-sectional and longitudinal, randomized and non-randomized, and had control groups.