Non-invasive health prediction from visually observable features

Fan Yi Khong; Tee Connie; Michael Kah Ong Goh; Li Pei Wong; Pin Shen Teh; Ai Ling Choo

doi:10.12688/f1000research.72894.2

Home Browse Non-invasive health prediction from visually observable features

ALL Metrics

Views

Downloads

Get PDF

Get XML

Export

▬

✚

Research Article

Revised

Non-invasive health prediction from visually observable features

[version 2; peer review: 1 approved, 2 approved with reservations]

Fan Yi Khong¹, Tee Connie ¹, Michael Kah Ong Goh¹, Li Pei Wong², Pin Shen Teh³, Ai Ling Choo⁴

Fan Yi Khong¹, Tee Connie ¹, [...] Michael Kah Ong Goh¹, Li Pei Wong², Pin Shen Teh³, Ai Ling Choo⁴

PUBLISHED 02 Mar 2022

Author details Author details

¹ Faculty of Information Science and Technology, Multimedia University, Melaka, Melaka, 75450, Malaysia
² School of Computer Sciences, Universiti Sains Malaysia, Penang, Penang, 11800, Malaysia
³ Department of Operations, Technology, Events and Hospitality Management, Faculty of Business and Law, Manchester Metropolitan University, Manchester, Manchester, M15 6BH, UK
⁴ iRadar Sdn. Bhd., Melaka, Melaka, 75450, Malaysia

Fan Yi Khong
Roles: Data Curation, Formal Analysis, Investigation, Methodology, Writing – Original Draft Preparation

Tee Connie
Roles: Conceptualization, Project Administration, Resources, Supervision, Validation, Writing – Review & Editing

Michael Kah Ong Goh
Roles: Investigation, Software, Validation, Writing – Review & Editing

Li Pei Wong
Roles: Formal Analysis, Validation, Visualization, Writing – Review & Editing

Pin Shen Teh
Roles: Validation, Visualization, Writing – Review & Editing

Ai Ling Choo
Roles: Validation, Writing – Review & Editing

OPEN PEER REVIEW

REVIEWER STATUS

This article is included in the Research Synergy Foundation gateway.

This article is included in the Artificial Intelligence and Machine Learning gateway.

Abstract

Background: The unprecedented development of Artificial Intelligence has revolutionised the healthcare industry. In the next generation of healthcare systems, self-diagnosis will be pivotal to personalised healthcare services. During the COVID-19 pandemic, new screening and diagnostic approaches like mobile health are well-positioned to reduce disease spread and overcome geographical barriers. This paper presents a non-invasive screening approach to predict the health of a person from visually observable features using machine learning techniques. Images like face and skin surface of the patients are acquired using camera or mobile devices and analysed to derive clinical reasoning and prediction of the person’s health.
Methods: In specific, a two-level classification approach is presented. The proposed hierarchical model chooses a class by training a binary classifier at the node of the hierarchy. Prediction is then made using a set of class-specific reduced feature set.
Results: Testing accuracies of 86.87% and 76.84% are reported for the first and second-level classification. Empirical results demonstrate that the proposed approach yields favourable prediction results while greatly reduces the computational time.
Conclusions: The study suggests that it is possible to predict the health condition of a person based on his/her face appearance using cost-effective machine learning approaches.

Keywords

Machine learning, Health prediction, Remote screening and diagnosis

Corresponding author: Tee Connie

Competing interests: No competing interests were disclosed.

Grant information: The author(s) declared that no grants were involved in supporting this work.

Copyright: © 2022 Khong FY et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite: Khong FY, Connie T, Goh MKO et al. Non-invasive health prediction from visually observable features [version 2; peer review: 1 approved, 2 approved with reservations]. F1000Research 2022, 10:918 (https://doi.org/10.12688/f1000research.72894.2) First published: 13 Sep 2021, 10:918 (https://doi.org/10.12688/f1000research.72894.1) Latest published: 02 Mar 2022, 10:918 (https://doi.org/10.12688/f1000research.72894.2)

Revised Amendments from Version 1

Following are the changes from previous version of the article:
- More work related to the study has been added in the Literature Review
- More details about how the features are extracted from the images are provided in the Proposed Solution
- A comparison with state-of-the-art methods has been presented in the Experiment
These changes are reflected in a revised Table 1 and new Table 6.

See the authors' detailed response to the review by Prabina Kumar Meher

Introduction

As technology advances, machine learning techniques have been growing in popularity over the past years. Machine learning techniques have proven to be effective in solving many modern problems in different domains. There is an increased research interest in applying machine learning methods for clinical informatics and healthcare systems.^1-4 Meanwhile, facial recognition technology has been vastly utilized in various fields. For instance, it has been applied to unlock phones, find wanted fugitives and diagnose diseases. There have been many kinds of research done on disease diagnosis using facial images.^5-8 Systems that only use facial features to diagnose illnesses are beneficial for remote medical diagnosis.

In this research, a machine learning approach was developed to detect the health condition of a person based on facial features. The purpose of the health prediction system was to identify images as ‘healthy’, or ‘ill’ with either ‘fever’, ‘sore throat’, or ‘runny nose’ symptoms. Facial images containing healthy and ill faces (fever, sore throat and runny nose) were collected. Then, discriminative facial features were extracted from the images using different feature extraction techniques. These features were used to train several machine learning classifiers for health prediction.

Literature review

In this section, various types of approaches to health prediction using facial features are studied and reviewed to learn about their respective advantages and disadvantages. These approaches are separated into two categories: conventional approaches and deep learning approaches.

Conventional approaches

In 2013, Zhao et al.¹ introduced an approach to classify Down Syndrome through image-based facial dysmorphology. Facial features were extracted using Contourlet transform-based and local binary pattern- based (LBP) local texture features, as well as geometric features using landmarks of facial anatomy. The support vector machine (SVM) classifier, this technique has produced an accuracy of 97.92%.

A survey done by² about genetic disorders diagnosis based on facial images, Saraydemir et al.³ presented an approach to identify subjects with Down Syndrome from healthy subjects using facial image. Gabor wavelet transform (GWT) was implemented for feature extraction purposes. Then, linear discriminant analysis (LDA) and principal component analysis (PCA) were carried out for the reduction of dimension. 96% and 97.34% accuracy were produced.

A research conducted by⁵ developed an approach for identifying Down Syndrome based on analysis of facial landmarks on 2D images. An independent component analysis-based hierarchical constrained local model (HCLM) was introduced to identify the landmarks of a face. The method was also tested on a mixed-syndromes dataset, and the highest accuracy achieved was 97%.

Another study related to health prediction systems using facial features that uses traditional machine learning methods, is an acromegaly identification using facial images proposed by.⁶ A few conventional methods such as SVM, generalized linear models (LM), k-NN, RF of randomized trees (RT) as well as other deep learning methods were used to train the model. The best performance was attained by the SVM method with a 95% PPV and 88% NPV, and with an accuracy of 91%. With frontalized faces, k-NN worked best with 89% PPV and 93% NPV, also with an accuracy of 91%.

Deep learning approaches

In 2018, Sajid et al.⁷ developed a palsy grading system based on unsymmetrical facial features using deep learning. A convolutional neural network (CNN) was proposed to extract features that exhibited palsy symptoms from a large number of facial images. The results of the model on the improved dataset showed a recognition rate of 92.6%.

A facial analysis framework introduced by⁸ called DeepGestalt, to identify rare genetic syndromes using deep learning. The training process of the DeepGestalt model consisted of two steps. Firstly, an overall representation of the face was learned by the model. The binary classification problem of identifying Angelman Syndrome (AS) and Cornelia De Lange Syndrome (CdLS) patients achieved an accuracy of 92% and 96.88%, respectively.

In year 2020,⁹ proposed to detect cancer using the facial features of patients. They used the network architecture of a residual neural network (ResNet) which comprised 27 convolution layers and two fully connected (FC) layers. Transfer learning was also applied for convolution layers 1-5 by directly obtaining the weights from a pre-trained face encoding model developed by.¹⁰ To describe the distinguishing traits of non-cancer and cancer datasets, they used gradient-weighted class activation mapping (grad-CAM) for the model that they trained. The accuracy rate produced by this approach was 82%.

Apart from that,¹¹ developed a technique to detect Down Syndrome automatically based on facial images with deep convolutional neural network (DCNN). Firstly, they trained a DCNN on a large dataset to acquire an overall face encoding network. The network architecture consists of ten convolutional layers activated by ReLU along with three FC layers. This method achieved an accuracy of 95.87%.

Also in 2020,¹² developed a study to diagnose and classify the severity of acromegaly at different severity levels using facial images with deep learning. CNN was used in this method. For facial recognition, the pre-trained Inception ResNet V1 was utilized to extract features. The total prediction accuracy achieved by this method was 90.7%.

Besides, Forte et al.¹³ presented a deep learning approach to assess a patient’s health by using facial and bodily cues. To increase the dataset size, a synthetic dataset containing acutely ill images were generated using a neural transfer CNN network. After that, four CNN models were trained on different parts of the faces and the features were concatenated into a final feature and fed to a staked CNN. The proposed model was tested using a dataset that was made up of images of volunteers injected with lipopolysaccharide.

On the other hand, Onyema et al.¹⁴ performed facial recognition for patients monitoring using ResNet. Facial emotions is believed to be closely related to the patient’s state of mind. The seven universal emotions including happy, sad, fear, anger, surprise and neural were investigated. Data augmentation was applied to increase the diversity of the data. An accuracy of 70% was achieved using the proposed approach.

Recently, Connie et al.¹⁵ proposed an explainable AI approach for providing explanations for the predictions made by an AI model for health application. A transfer learning approach with VGGFace model was applied to process the facial images. After that, an outcome whether the face belongs to a sick person was derived. Explainable AI (XAI) was used to provide explanation why the outcome, e.g. sick or healthy face, was produced. Different XAI techniques including Integrated Gradient, Explainable region-based AI (XRAI) and Local Interpretable Model-Agnostic Explanations (LIME) were investigated in the paper. The proposed approach had helped to increase the accountability of the healthcare system. A summary of works related to this study (hand-crafted features based methods), together with the pros and cons of each method, is presented in Table 1.

Table 1. A summary of works related to this study.

Author	Method	Database	Classes	Recognition Rate	Pros	Cons
Zhao, Q., Rosenbaum, K., Sze, R., Zand, D., Summar, M., & Linguraru, M. G.¹	▪ Geometric + SVM ▪ Texture + SVM ▪ Combined + SVM	Self-collected dataset	Down syndrome + Normal	97.92%	1. Contourlets preserve important wavelet features and provide a high level of anisotropy and directionality 2. LBP features are robust against illumination changes and takes less computational time	Facial anatomical landmarks and texture features need to be defined manually, requires more time and effort
Saraydemir, Ş., Taşpınar, N., Eroğul, O., Kayserili, H., & Dinçkan, N.³	▪ GWT + PCA & LDA + SVM ▪ GWT + PCA & LDA + k-NN	▪ University Medicine Faculty Department of Medical Genetics ▪ Down Syndrome Association of Turkey and Istanbul	Down syndrome + Healthy	97.34%	1. Dataset is small to produce robust results 2. Resistant to biases due to pose, illumination, and expression variances	Manual normalization requires more effort and time than automated approaches
Ferry, Q., Steinberg, J., Webber, C., FitzPatrick, D. R., Ponting, C. P., Zisserman, A., & Nellåker, C.⁴	PCA + AAM + k-NN	▪ Publicly available resources ▪ Scientifically published pictures of patients	Eight genetic disorders + Healthy	99.5%	1. Robust to artificial variations such as lighting, pose, and image quality 2. Provides consistent computational descriptions of facial gestalt	1. AAMs involve complex texture mapping and image warping operations which are susceptible to errors 2. AAMs have low performance on unseen faces
Zhao, Q., Okada, K., Rosenbaum, K., Kehoe, L., Zand, D. J., Sze, R., Summar, M., & Linguraru, M. G.⁵	Features: ▪ Geometric ▪ LBP ▪ Geometric + LBP ▪ GWT ▪ Geometric + GWT Classifiers: ▪ SVM-RBF ▪ Linear SVM ▪ k-NN ▪ RF ▪ LDA	Self-collected dataset	Down syndrome + Healthy	96.7%	1. CLMs are more generative and discriminative on unseen appearance 2. CLMs are more constant to global illumination variation and occlusion	1. ICA requires large datasets to train to produce good results 2. Optimization can converge to local minima or false locations
		Self-collected dataset	Mixed syndromes + Healthy	97%
Kong, X., Gong, S., Su, L., Howard, N., & Kong, Y.⁶	▪ k-NN ▪ SVM ▪ RF	▪ SCUT-FBP dataset ▪ Neurosurgery inpatient departments of hospitals in China ▪ Self-collected dataset	Acromegaly + Normal	95%	SVM performs well on extracted facial features	1. A possibility of bias caused by the selection of samples may occur 2. It is not known whether the outcome is generalizable to different populations

Methods

Proposed solution

A two-level classification approach is presented in this paper for health prediction based on facial features. Figure 1 shows the processes of how a prediction model was developed. First, facial images of healthy and ill (fever, sore throat and runny nose) persons were collected. Then, these images were pre-processed to clean, standardize and normalize the data. There are two levels of classification. The first-level classification is responsible for classifying samples into ‘healthy’ and ‘ill’ classes, while the second-level classification is in charge of classifying the ‘ill’ samples’ into ‘fever’, ‘sore throat’, and ‘runny nose’ classes. Therefore, there are two levels of model training in the proposed solution. In this research, conventional machine learning methods were adopted.

Figure 1. Proposed framework.

The feature extraction methods used were local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Gabor filter. LBP is a straightforward texture analysis method that constructs binary numbers by thresholding the neighbours of every pixel in an image. For every pixel, its eight neighbours are examined to see whether their intensity is higher than the particular pixel. The threshold results from the eight neighbours are used to construct an eight-digit binary number. If the intensity of the neighbour is less than or equal to the pixel, then the first digit of the binary number would be 0, otherwise, it would be 1. Then, the texture of the image is represented by a histogram of these numbers.

On the other hand, PCA is a dimensionality reduction method that works by finding out patterns and correlations that best represent the data in a least-square sense. Higher-dimensional data are projected to a lower-dimensional space. It is an unsupervised technique that does not consider labels. It seeks directions that maximize variance and are efficient for representation.

LDA is also a dimensionality reduction tool. Higher-dimensional data are projected to a lower-dimensional space. It works by finding the projection that best separates the data of two or more classes in a least-square sense. It is a supervised technique that seeks directions that maximizes the distance between classes and are efficient for discrimination.

On the contrary, Gabor filter is a technique used for texture analysis, edge detection, feature extraction and more. These filters have been claimed that they stimulate the visual system of some mammals. They can filter any particular frequencies in an image in the region of analysis. For example, they recognize some specific frequencies and ignore the rest. To analyse the texture from an image, a collection of Gabor filters containing different orientations are applied.

In this study, the pre-processed images are converted to grayscale before the features are extracted from the images. After that, the features from the images are extracted in two different ways. For LBP and Gabor filter features, the feature extraction procedure follows the order of: loading original and augmented images, extracting features from the whole dataset, separating the original and augmented images’ extracted features from the dataset, splitting the extracted features of the original dataset into training and testing sets, adding the extracted features of the augmented dataset into the training set, and shuffling, and scaling the training and testing sets (as required depending on the model’s performance).

On the other hand, for PCA and LDA features, the feature extraction procedure follows the order of: loading original and augmented images, splitting the original dataset into training and testing sets, adding the augmented dataset into the training set, and shuffling, scaling the training and testing sets (as required depending on the model’s performance), and extracting features from the training and testing sets.

The classifiers used were SVM, NN, KNN, and RF. A total of 16 combinations among the feature extraction techniques and classifiers mentioned were experimented with to find the best-performing model.

Datasets

In this study, a total of 733 facial images of healthy and ill persons were collected. Among 733 images, 233 are images of ill persons who had either fever, sore throat, or runny nose and 500 were images with healthy or normal persons. 420 out of the 500 healthy images contained normal faces of people from ages 1 to 50, while the remaining 80 images were healthy throat images. Images of healthy throat and ill persons were manually collected from various online sources, while images of healthy faces were obtained from the UTKFace database.¹⁶ The number of images for each class and subclass is listed in Table 2.

Table 2. Number of images for each class and subclass.

Class	Subclass	Number of images
Healthy	-	500
Il	Fever	78
	Sore throat	80
	Running nose	75

Results and discussion

In this section, the experimental results for the different models that consist of the combinations of four feature extraction methods and four classifiers are presented, analysed and discussed. The testing accuracies of the first and second-level classification of each model were recorded for 10 runs.¹⁷

SVM variants

The first experiment validates the performance of the SVM variant. Table 3 demonstrates the results of the SVM variant for the first and second classification tests. Among LBP, PCA, LDA and Gabor filter features, PCA features performed the best with SVM in the first-level classification. It achieves a promising result of 85.85% average testing accuracy with minimal overfitting. On the other hand, the LBP features performed the best with SVM in the second-level classification, with an average testing accuracy of 73.32%. The SVM variants generally produced results with the least overfitting among all the classifiers.¹⁸

Table 3. Experimental results of SVM variants.

Methods	1^st Level classification testing	2^nd Level classification testing
LBP + SVM	80.88	73.32
PCA + SVM	85.85	64.05
LDS + SVM	85.37	63.01
GABOR FILTER + SVM	81.29	63.45

NN variants

The experimental results for the NN variants are depicted in Table 6. Among all the feature extraction techniques, PCA features worked best with NN in the first-level classification. It achieved an average testing accuracy as high as 91.84%. On the other hand, the LBP features performed best with NN in the second-level classification with an average testing accuracy of 76.84%. In the second-level classification, the LBP model was also the only model that stood out among the other NN variants.¹⁹

KNN variants

The performance of the KNN variants for the first and second level classifications is given in Table 5. Among all the feature extraction techniques, again, PCA features worked best with KNN in the first-level classification, with an average testing accuracy as high as 90.34%. The same model also performed best in the second-level classification among all the KNN variants, as it obtained an average testing accuracy of 70.03%.

RF variants

The experimental results for the RF variants are displayed in Table 6. Among all the features extraction techniques, once again, at 88.57% average testing accuracy, PCA features performed the best with RF in the first-level classification. This model also scored best in the second-level classification among all the RF variants as it obtained an average testing accuracy of 74.15%.

First-level classification results

According to the experimental results of all the models shown in Tables 3 to 6, two models achieved over 90% average testing accuracies in the first-level classification. These models are the PCA+NN and PCA+KNN model.

Table 4. Experimental results of NN variants.

The best results are highlighted in bold.

Methods	1^st Level classification testing	2^nd Level classification testing
LBP + NN	86.87	76.84
PCA + NN	91.84	66.96
LDA + NN	86.87	63.19
GABOR FILTER + NN	86.05	66.93

Table 5. Experimental results of KNN variants.

The best results are highlighted in bold.

Methods	1^st Level classification testing	2^nd Level classification testing
LBP + NN	83.54	65.33
PCA + KNN	90.34	70.03
LDA + KNN	86.53	63.78
GABOR FILTER + KNN	72.26	63.02

Table 6. Experimental results of RF variants.

The best results are highlighted in bold.

Methods	1^st Level classification testing	2^nd Level classification testing
LBP + RF	83.61	67.95
PCA + RF	88.57	74.15
LDA + RF	85.31	64.15
GABOR FILTER + RF	87.89	62.41

PCA+NN

The model that achieved the highest accuracy in the first-level classification was PCA+NN. It obtained a 91.84% average testing accuracy. The high accuracy could be due to the fact that PCA effectively reduces the dimensions of data and it is able to capture the important correlations and patterns that best characterize the data. The misclassified samples were plotted during one of the runs of the finalized PCA+NN model. Out of the 147 samples, there were 15 misclassified samples.

PCA+KNN

The PCA+KNN model obtained the second-highest accuracies in the first-level classification after PCA+NN. Its performance was as good as that of PCA+KNN as it achieved a 90.34% average testing accuracy. Figure 2 shows the confusion matrix after running the first-level classification of PCA+KNN model. There is not much difference between the performance of PCA+NN and PCA+KNN as both of them were able to perform equally well.

Figure 2. Confusion matrix of PCA+KNN at first-level classification.

Overall first-level performance

Apart from PCA+NN and PCA+KNN, the overall results of the first-level classification were rather good as most of the models achieved an average testing accuracy of 80% and above. Even though the other models overfit more than PCA+NN and PCA+KNN, their results were still considered rather satisfactory. The symptoms shown on the faces of ill people or sore throats are important features to help the model classify healthy and ill samples.²⁰

Second-level classification results

Based on the results given in Tables 3 to 6, a total of four models achieved average testing accuracies between 70% and 77% in the second-level classification. These models were the LBP+NN, PCA+RF, LBP+SVM, and PCA+KNN model.

LBP+NN

The model that achieved the highest accuracy in the second-level classification was LBP+NN. It obtained an average testing accuracy of 76.84%. Its performance was considered rather satisfactory, as most of the other models only obtained testing accuracies between 60% and 68% on average. The reason that LBP+NN could perform well could be that the LBP features were invariant to illumination and were highly discriminative.

PCA+RF

The PCA+RF model performed nearly as well compared to LBP+NN with an average testing accuracy of 74.15% in the second-level classification. It performed well due to the previously mentioned benefits of the combination of PCA and RF being a classifier with outstanding predictive capabilities. Figure 3 shows the confusion matrix produced after running the second-level classification of PCA+RF model.

Figure 3. Confusion matrix of PCA+RF at second-level classification.

The confusion matrix was generated during one of the runs of the finalized PCA+RF model. The 0 label represents the ‘fever’ class, 1 represents the ‘sore throat’ class and 2 represents the ‘running nose’ class. It can be seen that the top two misclassified classes were the ‘fever’ (0) and ‘runny nose’ (2) classes, with 15 fever samples misclassified as runny nose and seven runny nose samples misclassified as fever. The reason for this occurrence is the same as for the LBP+NN model’s case. The total number of samples misclassified by PCA+RF for this run was 31 samples, with only one additional misclassified sample compared to the LBP+NN model. Hence, PCA+RF was able to produce results as good as LBP+NN in the second-level classification.

LBP+SVM and PCA+KNN

Other than the LBP+NN and PCA+RF, the two remaining models that achieved over 70% average testing accuracies were LBP+SVM and PCA+KNN. The LBP+SVM model obtained a 73.32% average testing accuracy in the second-level classification. The reason behind its performance is the robustness of LBP as well as the fact that SVM is effective in situations where the number of dimensions is larger than the number of samples. In the model’s second-level classification, the number of testing samples was always lesser than the number of dimensions.

Best model for the health prediction system

Among all models, the LBP+NN variant had the best overall performance in the first and second-level classifications. It achieved the highest average testing accuracy of 76.84% in the second-level classification. It also performed considerably well in the first-level classification with lesser overfitting than the other models with similar performances, as it showed 94.38% and 86.87% average training and testing accuracies, respectively.

Comparison with other methods

A comparison of the proposed methods with state-of-the-art approaches is presented in Table 7. It can be observed that the deep learning approaches including CNN¹ and VGGFace² outperform the proposed methods that rely on hand-crafted features. Nevertheless, the proposed approach has a great advantage as compared to the deep learning approaches in terms of computational speed. For example, it only took 0.0015 seconds to train the PCA+RF classifier, while it takes more than five minutes to perform training using the deep learning models. Therefore given a scenario where speed is a critical requirement and there is not many training samples available, the proposed method appears to be a more favourable choice.

Table 7. A comparison with state-of-the-art methods.

Methods	1 ^st Level classification testing
LBP + NN (First level classification)	86.87
PCA + RF (First level classification)	88.57
CNN⁷	92.71
VGGFace²¹	96.25

Conclusions

This paper presents a health prediction system using facial features evaluated using different machine learning models. Datasets containing facial images of healthy and ill (fever, sore throat and runny nose) persons were collected. The facial features of the images were extracted using LBP, PCA, LDA and Gabor filter feature extraction techniques. The features were trained using SVM, NN, KNN and RF classifiers. Among the 16 models, the LBP+NN model yielded the best overall performance for both the first and second-level classifications. It obtained average testing accuracies of 86.87% and 76.84% for the first and second-level classification, respectively.

Data availability

Underlying data

UTKFace Large Scale Face Dataset: https://susanqq.github.io/UTKFace/.

As it is impossible to obtain the consent for the face images retrieved from the UTKFace dataset, the images cannot be shared in this article.

Software availability

Source code available from: https://doi.org/10.5281/zenodo.5266406.²²

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).

Acknowledgements

The authors would like to thank the authors of UTKFace database for sharing the face images to be used in this study. This work has been approved by MMU Research Ethics Committee (Approval number: EA1442021).

References

1. Zhao Q, Rosenbaum K, Sze R, et al.: Down syndrome detection from facial photographs using machine learning techniques. Novak CL, Aylward S, Eds.; 2013; p. 867003.
2. Rai EL, Werghi MC, Al Muhairi N, et al.: Using facial images for the diagnosis of genetic syndromes: A survey. 2015 Int Conf Communications, Signal Processing, and Their Applications (ICCSPA’15). 2015; 1–6. Publisher Full Text
3. Saraydemir Ş, Taşpınar N, Eroğul O, et al.: Down Syndrome Diagnosis Based on Gabor Wavelet Transform. J Med Syst . 2012; 36(5): 3205–3213. PubMed Abstract | Publisher Full Text
4. Ferry Q, Steinberg J, Webber C, et al.: Diagnostically relevant facial gestalt information from ordinary photos. ELife . 2014; 3: e02020. PubMed Abstract | Publisher Full Text | Free Full Text
5. Zhao Q, Okada K, Rosenbaum K, et al.: Digital facial dysmorphology for genetic screening: Hierarchical constrained local model using ICA. Med Image Anal . 2014; 18(5): 699–710. PubMed Abstract | Publisher Full Text
6. Kong X, Gong S, Su L, et al.: Automatic Detection of Acromegaly From Facial Photographs Using Machine Learning Methods. EBioMedicine . 2018; 27: 94–102. PubMed Abstract | Publisher Full Text | Free Full Text
7. Sajid M, Shafique T, Baig M, et al.: Automatic Grading of Palsy Using Asymmetrical Facial Features: A Study Complemented by New Solutions. Symmetry . 2018; 10(7): 242. Publisher Full Text
8. Gurovich Y, Hanani Y, Bar O, et al.: DeepGestalt - Identifying Rare Genetic Syndromes Using Deep Learning. ArXiv:1801.07637 [Cs]. 2018. Publisher Full Text
9. Liang B, Yang N, He G, et al.: Identification of the Facial Features of Patients With Cancer: A Deep Learning–Based Pilot Study. J Med Internet Res . 2020; 22(4): e17234. PubMed Abstract | Publisher Full Text | Free Full Text
10. Schroff F, Kalenichenko D, Philbin J: FaceNet: A unified embedding for face recognition and clustering. IEEE Conf Computer Vision Pattern Recognition (CVPR) . 2015; 2015: 815–823. Publisher Full Text
11. Qin B, Liang L, Wu J, et al.: Automatic Identification of Down Syndrome Using Facial Images with Deep Convolutional Neural Network. Diagnostics . 2020; 10(7): 487. PubMed Abstract | Publisher Full Text | Free Full Text
12. Kong Y, Kong X, He C, et al.: Constructing an automatic diagnosis and severity-classification model for acromegaly using facial photographs by deep learning. J Hematol Oncol . 2020; 13(1): 88. Publisher Full Text
13. Forte C, Voinea A, Chichirau M, et al.: Deep Learning for Identification of Acute Illness and Facial Cues of Illness. Front Med. 2021; 8: 661309. PubMed Abstract | Publisher Full Text
14. Onyema EM, Shukla PK, Dalal S, et al.: Enhancement of Patient Facial Recognition through Deep Learning Algorithm: ConvNet.J Healthc Eng.2021; 2021: 8. Publisher Full Text PubMed Abstract |
15. Connie T, Tan YF, Goh MKO, et al.: Explainable Health Prediction from Facial Features with Transfer Learning. J Intell Fuzzy Syst. . 2022; 42(3): 2491–2503. Publisher Full Text
16. Zhang Z, Song Y, Qi H: Age Progression/Regression by Conditional Adversarial Autoencoder. IEEE Conf Computer Vision Pattern Recognition (CVPR) . 2017: 4352–4360.
17. Jordan J: Evaluating a machine learning model.2017. Reference Source
18. Abdiansah A, Wardoyo R: Time Complexity Analysis of Support Vector Machines (SVM) in LibSVM. Int J Computer Applications . 2015; 128(3): 28–34. Publisher Full Text
19. Karis NS, Rafiqah N, Nursabillilah, et al.: Local Binary Pattern (LBP) with application to variant object detection: A survey and method. 12th Int Colloquium on Signal Processing & Its Applications (CSPA). 2016). 2016. Publisher Full Text
20. British Columbia Health Link BC: Facts about Influenza.2021. Reference Source
21. Parkhi OM, Vedaldi A, Zisserman A: Deep Face Recognition.Procedings of the British Machine Vision Conference 2015. British Machine Vision Association, Swansea. 2015: 41.1–41.12. Publisher Full Text
22. Khong FY, Connie T, Michael GKO, et al.: gkomix88/HealthPrediction: Non-invasive Health Prediction from Visually Observable Features (HealthPrediction). Zenodo. 2021. Publisher Full Text

Comments on this article Comments (0)

Version 2

VERSION 2 PUBLISHED 13 Sep 2021

Author details Author details

Fan Yi Khong
Roles: Data Curation, Formal Analysis, Investigation, Methodology, Writing – Original Draft Preparation

Tee Connie
Roles: Conceptualization, Project Administration, Resources, Supervision, Validation, Writing – Review & Editing

Michael Kah Ong Goh
Roles: Investigation, Software, Validation, Writing – Review & Editing

Li Pei Wong
Roles: Formal Analysis, Validation, Visualization, Writing – Review & Editing

Pin Shen Teh
Roles: Validation, Visualization, Writing – Review & Editing

Ai Ling Choo
Roles: Validation, Writing – Review & Editing

Competing interests

No competing interests were disclosed.

Grant information

The author(s) declared that no grants were involved in supporting this work.

Article Versions (2)

version 2

Revised

Published: 02 Mar 2022, 10:918

https://doi.org/10.12688/f1000research.72894.2

version 1

Published: 13 Sep 2021, 10:918

https://doi.org/10.12688/f1000research.72894.1

© 2022 Khong FY et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Download

Export To

metrics

	Views	Downloads
F1000Research	-	-
PubMed Central Data from PMC are received and updated monthly.	-	-

Citations

SEE MORE DETAILS

CITE

how to cite this article

Khong FY, Connie T, Goh MKO et al. Non-invasive health prediction from visually observable features [version 2; peer review: 1 approved, 2 approved with reservations]. F1000Research 2022, 10:918 (https://doi.org/10.12688/f1000research.72894.2)

NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.

track

receive updates on this article

Track an article to receive email alerts on any updates to this article.

Open Peer Review

Current Reviewer Status: ?

Key to Reviewer Statuses VIEW HIDE

ApprovedThe paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approvedFundamental flaws in the paper seriously undermine the findings and conclusions

Version 2

VERSION 2

PUBLISHED 02 Mar 2022

Revised

Views

Reviewer Report 25 Apr 2022

Yong Wee Sek, Fakulti Teknologi Maklumat dan Komunikasi, Universiti Teknikal Malaysia Melaka, Melaka, Malaysia

Approved with Reservations

https://doi.org/10.5256/f1000research.121466.r134514

In this paper, the authors introduce a machine learning approach to detect the health condition of a person based on facial features. There is a great deal of good work in this paper. However, it does require a significant overall revision to make the best of this and present the work in a clear and well-structured way, particularly in the introduction, literature review, methodology, results, and discussions in order to achieve the research objectives. Additionally, the discussion section is not exhaustive enough to provide a good attempt to explain many of the study results.

Abstract

Background – This section should discuss the existing problems of self-diagnosis in the health care system. Why non-invasive screening approach proposed in this study? How this approach is able to overcome the existing self-diagnosis problems? Please include the main objective(s) or research question(s) of this study in this section. The proposed machine learning techniques have to be specifically stated in the methods section. The results section should discuss the result of the combination techniques for each level of classification.

Introduction

This section should include the following information:

What are the existing problems in self-diagnosis in the health care system?
Why is a non-invasive screening approach proposed in this research?
How is the non-invasive screening approach proposed can overcome these problems?
What are the existing non-invasive screening approach that has been proposed to solve existing problems in self-diagnosis in the health care system?
What are the weaknesses of existing methods?
How is the proposed non-invasive screening approach, able to solve the existing problems in self-diagnosis in the health care system?
What are the techniques that have been proposed to extract facial features, especially for the health care system? The weaknesses of existing techniques. How is the proposed technique can overcome the current weaknesses?
Please include the main objective(s)/research questions of this study in this section.
Include a brief description of the methodology and classification methods used in this study.

Literature review/Background study

In this section, discussion on the following should be included:

The major problems in self-diagnosis in healthcare system
Why non-invasive screening approach suitable to overcome the self-diagnosis problem in the health care system?
What are the existing methods proposed to overcome these problems? Should discuss the advantages and disadvantages of the existing methods.
Why are machine learning approaches introduced in solving self-diagnosis problems in the health care system?
Discuss the advantages of proposing a machine learning approach in solving self-diagnosis problems in the health care system

Methodology

•Discussion about what are the outputs for each method in each classification level. What are the features being extracted? How are these outputs used from one level to another level?
Discussion about how to obtain the dataset? How are the training and testing datasets divided?

Results

This section should present the following results:

without combination methods/ single method
with combination methods

Discussions

The author(s) should discuss/relate/compare the results with prior relevant studies. Explanation how the performance of the proposed techniques. Provide justification for the performance.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Partly
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Machine learning, information system, AI

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

CITE

Report a concern

Respond or Comment

Version 1

VERSION 1

PUBLISHED 13 Sep 2021

Views

Reviewer Report 14 Feb 2022

Prabina Kumar Meher, ICAR-Indian Agricultural Statistics Research Institute, New Delhi, Delhi, India

Approved with Reservations

https://doi.org/10.5256/f1000research.76503.r121344

The authors developed a machine learning-based approach for predicting the health status of an individual by using the image data of their face. Below are my comments for further improvement of the article.

The authors must

The authors must clearly describe how the features were generated from the image dataset.
The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.
The author should perform a comparative analysis with the existing methods to claim the superiority of the method.
The author must try to establish an online prediction tool for the real use of the developed approach.
More work related to this subject must be discussed.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

No
If applicable, is the statistical analysis and its interpretation appropriate?

Partly
Are all the source data underlying the results available to ensure full reproducibility?

Partly
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Application of Machine learning for prediction/classification using biological data in agricultural and biomedical science.

CITE

Report a concern

Author Response 02 Mar 2022

Tee Connie, Faculty of Information Science and Technology, Multimedia University, Melaka, 75450, Malaysia

02 Mar 2022

Author Response

Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, ... Continue reading Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, more analysis has been conducted. In this response letter, we list the specific concerns and questions raised by the reviewer and provide our itemized response.

Point 1: The authors must clearly describe how the features were generated from the image dataset.

Response 1: We thank the reviewer for the careful review and comment. We understand that it is important to describe how the features were generated from the images. In the study, four feature extraction methods namely local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Gabor filter were applied. The details how the features are extracted from the image have been added.

Point 2: The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.

Response 2: This is a very good suggestion. In fact, we had also performed experiments using CNN on the image data. However, as the scope of the paper is more towards the analysis and comparison of conventional feature extraction and classification methods rather than the deep learning approach, we did not include the details for CNN in the paper.

Anyway, to better illustrate the use of CNN on the health dataset, we provide some information about the experiments using CNN in this response letter here.

Point 3: The author should perform a comparative analysis with the existing methods to claim the superiority of the method.

Response 3: We thank the reviewer for the suggestion. It is important to provide a comparative analysis with the existing methods to claim the superiority of the method. To better illustrate the performance of the proposed methods as compared to state-of-the-art techniques, we have added a new section in Experiments.

Point 4: The author must try to establish an online prediction tool for the real use of the developed approach.

Response 4: We thank the for the suggestion. It is advantageous to establish an online prediction tool for the real use of the developed approach and we will consider this in our future endeavors.

Point 5: More work related to this subject must be discussed.

Response 5: Thank you for the suggestion. In addition to the related works that have been discussed in the Literature Review section in the paper, more studies related to the use of face images for health prediction are provided in the article.
Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, more analysis has been conducted. In this response letter, we list the specific concerns and questions raised by the reviewer and provide our itemized response.

Point 1: The authors must clearly describe how the features were generated from the image dataset.

Response 1: We thank the reviewer for the careful review and comment. We understand that it is important to describe how the features were generated from the images. In the study, four feature extraction methods namely local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Gabor filter were applied. The details how the features are extracted from the image have been added.

Point 2: The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.

Response 2: This is a very good suggestion. In fact, we had also performed experiments using CNN on the image data. However, as the scope of the paper is more towards the analysis and comparison of conventional feature extraction and classification methods rather than the deep learning approach, we did not include the details for CNN in the paper.

Anyway, to better illustrate the use of CNN on the health dataset, we provide some information about the experiments using CNN in this response letter here.

Point 3: The author should perform a comparative analysis with the existing methods to claim the superiority of the method.

Response 3: We thank the reviewer for the suggestion. It is important to provide a comparative analysis with the existing methods to claim the superiority of the method. To better illustrate the performance of the proposed methods as compared to state-of-the-art techniques, we have added a new section in Experiments.

Point 4: The author must try to establish an online prediction tool for the real use of the developed approach.

Response 4: We thank the for the suggestion. It is advantageous to establish an online prediction tool for the real use of the developed approach and we will consider this in our future endeavors.

Point 5: More work related to this subject must be discussed.

Response 5: Thank you for the suggestion. In addition to the related works that have been discussed in the Literature Review section in the paper, more studies related to the use of face images for health prediction are provided in the article.
Competing Interests: No competing interests were disclosed. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 02 Mar 2022

Tee Connie, Faculty of Information Science and Technology, Multimedia University, Melaka, 75450, Malaysia

02 Mar 2022

Author Response

Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, ... Continue reading Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, more analysis has been conducted. In this response letter, we list the specific concerns and questions raised by the reviewer and provide our itemized response.

Point 1: The authors must clearly describe how the features were generated from the image dataset.

Response 1: We thank the reviewer for the careful review and comment. We understand that it is important to describe how the features were generated from the images. In the study, four feature extraction methods namely local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Gabor filter were applied. The details how the features are extracted from the image have been added.

Point 2: The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.

Response 2: This is a very good suggestion. In fact, we had also performed experiments using CNN on the image data. However, as the scope of the paper is more towards the analysis and comparison of conventional feature extraction and classification methods rather than the deep learning approach, we did not include the details for CNN in the paper.

Anyway, to better illustrate the use of CNN on the health dataset, we provide some information about the experiments using CNN in this response letter here.

Point 3: The author should perform a comparative analysis with the existing methods to claim the superiority of the method.

Response 3: We thank the reviewer for the suggestion. It is important to provide a comparative analysis with the existing methods to claim the superiority of the method. To better illustrate the performance of the proposed methods as compared to state-of-the-art techniques, we have added a new section in Experiments.

Point 4: The author must try to establish an online prediction tool for the real use of the developed approach.

Response 4: We thank the for the suggestion. It is advantageous to establish an online prediction tool for the real use of the developed approach and we will consider this in our future endeavors.

Point 5: More work related to this subject must be discussed.

Response 5: Thank you for the suggestion. In addition to the related works that have been discussed in the Literature Review section in the paper, more studies related to the use of face images for health prediction are provided in the article.
Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, more analysis has been conducted. In this response letter, we list the specific concerns and questions raised by the reviewer and provide our itemized response.

Point 1: The authors must clearly describe how the features were generated from the image dataset.

Response 1: We thank the reviewer for the careful review and comment. We understand that it is important to describe how the features were generated from the images. In the study, four feature extraction methods namely local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Gabor filter were applied. The details how the features are extracted from the image have been added.

Point 2: The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.

Response 2: This is a very good suggestion. In fact, we had also performed experiments using CNN on the image data. However, as the scope of the paper is more towards the analysis and comparison of conventional feature extraction and classification methods rather than the deep learning approach, we did not include the details for CNN in the paper.

Anyway, to better illustrate the use of CNN on the health dataset, we provide some information about the experiments using CNN in this response letter here.

Point 3: The author should perform a comparative analysis with the existing methods to claim the superiority of the method.

Response 3: We thank the reviewer for the suggestion. It is important to provide a comparative analysis with the existing methods to claim the superiority of the method. To better illustrate the performance of the proposed methods as compared to state-of-the-art techniques, we have added a new section in Experiments.

Point 4: The author must try to establish an online prediction tool for the real use of the developed approach.

Response 4: We thank the for the suggestion. It is advantageous to establish an online prediction tool for the real use of the developed approach and we will consider this in our future endeavors.

Point 5: More work related to this subject must be discussed.

Response 5: Thank you for the suggestion. In addition to the related works that have been discussed in the Literature Review section in the paper, more studies related to the use of face images for health prediction are provided in the article.
Competing Interests: No competing interests were disclosed. Close
Report a concern

Views

Reviewer Report 14 Sep 2021

Andrews Samraj, Department of Computer Science and Engineering, Mahendra Engineering College, Namakkal, Tamil Nadu, India; Mahendra Engineering College, Namakkal, Tamil Nadu, India

Approved

https://doi.org/10.5256/f1000research.76503.r94202

Non-invasive health prediction is the essential need of the hour due to the pandemic situation in health care. The approach and modeling using image data set of human faces with sound health and with illness are taken for research and the findings were presented in this work.

Though the testing accuracies are low, an increase in the number of samples may help in improving.

The modifications suggested:

In methods, the two-level classification has to be depicted clearly.
Tables are not visible. Improve font size
Results and discussion should only explain the results, assumptions, and parameters. So explain the variants in the Methods section.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Partly
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Image processing, Non invasive signal and image analysis, Soft cyborgs, Wearable Technology, Bionics, AI-ML

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

CITE

Report a concern

Respond or Comment

Comments on this article Comments (0)

Version 2

VERSION 2 PUBLISHED 13 Sep 2021

Open Peer Review

Reviewer Status

Reviewer Reports

	Invited Reviewers
	1	2	3
Version 2 (revision) 02 Mar 22			read
Version 1 13 Sep 21	read	read

Andrews Samraj, Mahendra Engineering College, Namakkal, India; Mahendra Engineering College, Namakkal, India
Prabina Kumar Meher, ICAR-Indian Agricultural Statistics Research Institute, New Delhi, India
Yong Wee Sek, Universiti Teknikal Malaysia Melaka, Melaka, Malaysia

Comments on this article

All Comments(0)

Add a comment

Browse by related subjects

Back to all reports

Reviewer Report

5 Views

25 Apr 2022 | for Version 2

Yong Wee Sek, Fakulti Teknologi Maklumat dan Komunikasi, Universiti Teknikal Malaysia Melaka, Melaka, Malaysia

5 Views Cite this report Responses(0)

Approved With Reservations

What are the existing problems in self-diagnosis in the health care system?
Why is a non-invasive screening approach proposed in this research?
How is the non-invasive screening approach proposed can overcome these problems?
What are the existing non-invasive screening approach that has been proposed to solve existing problems in self-diagnosis in the health care system?
What are the weaknesses of existing methods?
How is the proposed non-invasive screening approach, able to solve the existing problems in self-diagnosis in the health care system?
What are the techniques that have been proposed to extract facial features, especially for the health care system? The weaknesses of existing techniques. How is the proposed technique can overcome the current weaknesses?
Please include the main objective(s)/research questions of this study in this section.
Include a brief description of the methodology and classification methods used in this study.

Literature review/Background study

In this section, discussion on the following should be included:

The major problems in self-diagnosis in healthcare system
Why non-invasive screening approach suitable to overcome the self-diagnosis problem in the health care system?
What are the existing methods proposed to overcome these problems? Should discuss the advantages and disadvantages of the existing methods.
Why are machine learning approaches introduced in solving self-diagnosis problems in the health care system?
Discuss the advantages of proposing a machine learning approach in solving self-diagnosis problems in the health care system

Methodology

•Discussion about what are the outputs for each method in each classification level. What are the features being extracted? How are these outputs used from one level to another level?
Discussion about how to obtain the dataset? How are the training and testing datasets divided?

Results

This section should present the following results:

without combination methods/ single method
with combination methods

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Partly
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Machine learning, information system, AI

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

18 Views

14 Feb 2022 | for Version 1

Prabina Kumar Meher, ICAR-Indian Agricultural Statistics Research Institute, New Delhi, Delhi, India

18 Views Cite this report Responses(1)

Approved With Reservations

The authors must clearly describe how the features were generated from the image dataset.
The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.
The author should perform a comparative analysis with the existing methods to claim the superiority of the method.
The author must try to establish an online prediction tool for the real use of the developed approach.
More work related to this subject must be discussed.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

No
If applicable, is the statistical analysis and its interpretation appropriate?

Partly
Are all the source data underlying the results available to ensure full reproducibility?

Partly
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Application of Machine learning for prediction/classification using biological data in agricultural and biomedical science.

Respond to this report

Responses (1)

Author Response

02 Mar 2022

Tee Connie, Faculty of Information Science and Technology, Multimedia University, Melaka, 75450, Malaysia

Dear Reviewer,

Thank you very much for your time and efforts in reviewing our manuscript “Non-Invasive Health Prediction from Visually Observable Features”. According to your valuable comments and suggestions, more analysis has been conducted. In this response letter, we list the specific concerns and questions raised by the reviewer and provide our itemized response.

Point 1: The authors must clearly describe how the features were generated from the image dataset.

Response 1: We thank the reviewer for the careful review and comment. We understand that it is important to describe how the features were generated from the images. In the study, four feature extraction methods namely local binary pattern (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Gabor filter were applied. The details how the features are extracted from the image have been added.

Point 2: The deep learning method convolutional neural network (CNN) is one of the most appropriate methods for prediction using image data. The authors should employ the CNN for the same and compare the accuracy with that of machine learning algorithms.

Response 2: This is a very good suggestion. In fact, we had also performed experiments using CNN on the image data. However, as the scope of the paper is more towards the analysis and comparison of conventional feature extraction and classification methods rather than the deep learning approach, we did not include the details for CNN in the paper.

Anyway, to better illustrate the use of CNN on the health dataset, we provide some information about the experiments using CNN in this response letter here.

Point 3: The author should perform a comparative analysis with the existing methods to claim the superiority of the method.

Response 3: We thank the reviewer for the suggestion. It is important to provide a comparative analysis with the existing methods to claim the superiority of the method. To better illustrate the performance of the proposed methods as compared to state-of-the-art techniques, we have added a new section in Experiments.

Point 4: The author must try to establish an online prediction tool for the real use of the developed approach.

Response 4: We thank the for the suggestion. It is advantageous to establish an online prediction tool for the real use of the developed approach and we will consider this in our future endeavors.

Point 5: More work related to this subject must be discussed.

Response 5: Thank you for the suggestion. In addition to the related works that have been discussed in the Literature Review section in the paper, more studies related to the use of face images for health prediction are provided in the article.

View more View less

Competing Interests

No competing interests were disclosed.

Back to all reports

Reviewer Report

30 Views

14 Sep 2021 | for Version 1

Andrews Samraj, Department of Computer Science and Engineering, Mahendra Engineering College, Namakkal, Tamil Nadu, India; Mahendra Engineering College, Namakkal, Tamil Nadu, India

30 Views Cite this report Responses(0)

Approved

In methods, the two-level classification has to be depicted clearly.
Tables are not visible. Improve font size
Results and discussion should only explain the results, assumptions, and parameters. So explain the variants in the Methods section.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Not applicable
Are all the source data underlying the results available to ensure full reproducibility?

Partly
Are the conclusions drawn adequately supported by the results?

Yes

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Image processing, Non invasive signal and image analysis, Soft cyborgs, Wearable Technology, Bionics, AI-ML

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Respond to this report

Responses (0)

Alongside their report, reviewers assign a status to the article:

Approved - the paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations - A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approved - fundamental flaws in the paper seriously undermine the findings and conclusions

[1] 1. Zhao Q, Rosenbaum K, Sze R, et al.: Down syndrome detection from facial photographs using machine learning techniques. Novak CL, Aylward S, Eds.; 2013; p. 867003.

[2] 2. Rai EL, Werghi MC, Al Muhairi N, et al.: Using facial images for the diagnosis of genetic syndromes: A survey. 2015 Int Conf Communications, Signal Processing, and Their Applications (ICCSPA’15). 2015; 1–6. Publisher Full Text

[3] 3. Saraydemir Ş, Taşpınar N, Eroğul O, et al.: Down Syndrome Diagnosis Based on Gabor Wavelet Transform. J Med Syst . 2012; 36(5): 3205–3213. PubMed Abstract | Publisher Full Text

[4] 4. Ferry Q, Steinberg J, Webber C, et al.: Diagnostically relevant facial gestalt information from ordinary photos. ELife . 2014; 3: e02020. PubMed Abstract | Publisher Full Text | Free Full Text

[5] 5. Zhao Q, Okada K, Rosenbaum K, et al.: Digital facial dysmorphology for genetic screening: Hierarchical constrained local model using ICA. Med Image Anal . 2014; 18(5): 699–710. PubMed Abstract | Publisher Full Text

[6] 6. Kong X, Gong S, Su L, et al.: Automatic Detection of Acromegaly From Facial Photographs Using Machine Learning Methods. EBioMedicine . 2018; 27: 94–102. PubMed Abstract | Publisher Full Text | Free Full Text

[7] 7. Sajid M, Shafique T, Baig M, et al.: Automatic Grading of Palsy Using Asymmetrical Facial Features: A Study Complemented by New Solutions. Symmetry . 2018; 10(7): 242. Publisher Full Text

[8] 8. Gurovich Y, Hanani Y, Bar O, et al.: DeepGestalt - Identifying Rare Genetic Syndromes Using Deep Learning. ArXiv:1801.07637 [Cs]. 2018. Publisher Full Text

[9] 9. Liang B, Yang N, He G, et al.: Identification of the Facial Features of Patients With Cancer: A Deep Learning–Based Pilot Study. J Med Internet Res . 2020; 22(4): e17234. PubMed Abstract | Publisher Full Text | Free Full Text

[10] 10. Schroff F, Kalenichenko D, Philbin J: FaceNet: A unified embedding for face recognition and clustering. IEEE Conf Computer Vision Pattern Recognition (CVPR) . 2015; 2015: 815–823. Publisher Full Text

[11] 11. Qin B, Liang L, Wu J, et al.: Automatic Identification of Down Syndrome Using Facial Images with Deep Convolutional Neural Network. Diagnostics . 2020; 10(7): 487. PubMed Abstract | Publisher Full Text | Free Full Text

[12] 12. Kong Y, Kong X, He C, et al.: Constructing an automatic diagnosis and severity-classification model for acromegaly using facial photographs by deep learning. J Hematol Oncol . 2020; 13(1): 88. Publisher Full Text

[13] 13. Forte C, Voinea A, Chichirau M, et al.: Deep Learning for Identification of Acute Illness and Facial Cues of Illness. Front Med. 2021; 8: 661309. PubMed Abstract | Publisher Full Text

[14] 14. Onyema EM, Shukla PK, Dalal S, et al.: Enhancement of Patient Facial Recognition through Deep Learning Algorithm: ConvNet.J Healthc Eng.2021; 2021: 8. Publisher Full Text PubMed Abstract |

[15] 15. Connie T, Tan YF, Goh MKO, et al.: Explainable Health Prediction from Facial Features with Transfer Learning. J Intell Fuzzy Syst. . 2022; 42(3): 2491–2503. Publisher Full Text

[16] 16. Zhang Z, Song Y, Qi H: Age Progression/Regression by Conditional Adversarial Autoencoder. IEEE Conf Computer Vision Pattern Recognition (CVPR) . 2017: 4352–4360.

[17] 17. Jordan J: Evaluating a machine learning model.2017. Reference Source

[18] 18. Abdiansah A, Wardoyo R: Time Complexity Analysis of Support Vector Machines (SVM) in LibSVM. Int J Computer Applications . 2015; 128(3): 28–34. Publisher Full Text

[19] 19. Karis NS, Rafiqah N, Nursabillilah, et al.: Local Binary Pattern (LBP) with application to variant object detection: A survey and method. 12th Int Colloquium on Signal Processing & Its Applications (CSPA). 2016). 2016. Publisher Full Text

[20] 20. British Columbia Health Link BC: Facts about Influenza.2021. Reference Source

[21] 21. Parkhi OM, Vedaldi A, Zisserman A: Deep Face Recognition.Procedings of the British Machine Vision Conference 2015. British Machine Vision Association, Swansea. 2015: 41.1–41.12. Publisher Full Text

[22] 22. Khong FY, Connie T, Michael GKO, et al.: gkomix88/HealthPrediction: Non-invasive Health Prediction from Visually Observable Features (HealthPrediction). Zenodo. 2021. Publisher Full Text

Non-invasive health prediction from visually observable features

Abstract

Keywords

Revised Amendments from Version 1

Introduction

Literature review

Table 1. A summary of works related to this study.

Methods

Proposed solution

Figure 1. Proposed framework.

Table 2. Number of images for each class and subclass.

Results and discussion

Table 3. Experimental results of SVM variants.

Table 4. Experimental results of NN variants.

Table 5. Experimental results of KNN variants.

Table 6. Experimental results of RF variants.

Figure 2. Confusion matrix of PCA+KNN at first-level classification.

Figure 3. Confusion matrix of PCA+RF at second-level classification.

Table 7. A comparison with state-of-the-art methods.

Conclusions

Data availability

Software availability

Acknowledgements

References

Comments on this article Comments (0)

Open Peer Review

Comments on this article Comments (0)

Open Peer Review

Reviewer Status

Reviewer Reports

Comments on this article

Browse by related subjects

Competing Interests Policy

Stay Updated