CNN-Based Facial Image Analysis for Pediatric Down Syndrome Classification

Yunidar Yunidar; Inda Mariana Harahap; Melinda Melinda; Rosmawinda Rosmawinda; Nurlida Basir; Aufa Rafiki; Imam Fathur Rahman

doi:10.35882/jeeemi.v8i2.1523

Yunidar Yunidar Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia https://orcid.org/0009-0008-4453-9215
Inda Mariana Harahap Department of Pediatry, Faculty of Nursing, Universitas Syiah Kuala, Darussalam, 23111, Banda Aceh, Indonesia https://orcid.org/0009-0004-4851-8666
Melinda Melinda Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia https://orcid.org/0000-0001-9082-6639
Rosmawinda Rosmawinda Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia https://orcid.org/0009-0004-8358-7691
Nurlida Basir Faculty of Science and Technology, Universiti Sains Islam Malaysia (USIM), Nilai Negeri Sembilan, Malaysia https://orcid.org/0000-0003-4738-1763
Aufa Rafiki Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia https://orcid.org/0009-0008-3031-8008
Imam Fathur Rahman Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia https://orcid.org/0009-0008-8068-994X

DOI: https://doi.org/10.35882/jeeemi.v8i2.1523

Keywords: Down Syndrome, CNN, Grad-CAM, Facial Landmark, Haar Cascade, Early Stopping, ANOVA

Abstract

Down syndrome (trisomy 21) is a genetic disorder caused by an extra copy of chromosome 21, resulting in distinctive developmental facial characteristics and intellectual delays. Early detection is crucial to enable timely medical intervention. However, conventional diagnostic procedures still rely on clinical observation and genetic testing, which can be invasive and expensive. This study proposes a facial image–based classification system for detecting Down syndrome using a Convolutional Neural Network (CNN) approach. Seven CNN architectures were evaluated, namely EfficientNetB0, MobileNetV2, ResNet34, ShuffleNetV2, AlexNet, VGG19, and InceptionV3, under two training scenarios: with and without early stopping. The dataset consisted of 1,000 facial images of children with and without Down syndrome, split into training, validation, and test sets at 60:20:20. Face detection was performed using the Haar Cascade Classifier, followed by data augmentation techniques including rotation, zoom, translation, horizontal flipping, and Gaussian noise to improve model generalization and reduce overfitting. Experimental results show that the VGG19 architecture achieved the best performance, with an accuracy of 94.5%, precision of 91.59%, recall of 98%, and an F1-score of 94.69%. A one-way ANOVA test yielded an F-value of 0.003 and a p-value of 0.955 (> 0.05), indicating no statistically significant difference between models trained with and without early stopping. Grad-CAM visualization highlighted key facial regions, namely the eyes, nose, and mouth, as the primary contributors to classification, while analysis using 68 facial landmark points revealed distinctive morphological patterns associated with Down syndrome. The integration of CNN models, Grad-CAM visualization, and facial landmark analysis demonstrates a promising, interpretable, and non-invasive approach to supporting early Down syndrome screening using facial images

Downloads

Download data is not yet available.

References

L. R. Chapman, I. V. P. Ramnarine, D. Zemke, A. Majid, and S. M. Bell, “Gene Expression Studies in Down Syndrome: What Do They Tell Us about Disease Phenotypes?,” Int. J. Mol. Sci., vol. 25, no. 5, 2024, doi: 10.3390/ijms25052968.

S. E. Antonarakis et al., “Down syndrome,” Nat. Rev. Dis. Prim., vol. 6, no. 1, pp. 1–43, 2021, doi: 10.1038/s41572-019-0143-7.

Word Population Review, “Down Syndrome Rates by Country 2025.” Accessed: Sep. 13, 2025. [Online]. Available: https://worldpopulationreview.com/country-rankings/down-syndrome-rates-by-country

T. I. W. Agustini Utari, Ferdy Kurniawan Cayami, Tithasiri Audi Rahardjo, Selvia Eva Sabatini, Vynda Ulvyana, “Critical Issue In The Identification Of Down Syndrome and Its Problems in Central Java, Indonesia: The Fact Of Needing Health Care And Better Management,” Intractable Rare Dis. Res., vol. 13, no. 2, pp. 121–125, 2024, doi: 10.5582/irdr.2023.01103.

K. R. Nattariya Vorravanpreecha, Thanayoot Lertboonnum, Rungrote Rodjanadit, Pak Sriplienchan, “Studying Down syndrome recognition probabilities in Thai children with de-identified computer-aided facial analysis,” Am. J. Med. Genet., vol. 176, no. 9, 2018, doi: https://doi.org/10.1002/ajmg.a.40483.

M. A. Shaikh and H. S. Al-Rawashdeh, “Deep Learning-Enabled Interpretable Down Syndrome Detection Model,” J. Disabil. Res., vol. 4, no. 3, pp. 1–12, 2025, doi: 10.57197/JDR-2025-0011.

A. Mumuni and F. Mumuni, “Data augmentation with automated machine learning: approaches and performance comparison with classical data augmentation methods,” Knowl. Inf. Syst., vol. 67, no. 5, pp. 4035–4085, 2025, doi: 10.1007/s10115-025-02349-x.

S. P. Arjunan and M. C. Thomas, “A Review of Ultrasound Imaging Techniques for the Detection of Down Syndrome,” Irbm, vol. 41, no. 2, pp. 115–123, 2020, doi: 10.1016/j.irbm.2019.10.004.

Vincy Devi VK and Rajesh R, “Down Syndrome Identification and Classification Using Facial Features With Neural Network,” Glob. J. Eng. Technol. Adv., vol. 12, no. 1, pp. 001–011, 2022, doi: 10.30574/gjeta.2022.12.1.0090.

A. Raza, K. Munir, M. S. Almutairi, and R. Sehar, “Novel Transfer Learning Based Deep Features for Diagnosis of Down Syndrome in Children Using Facial Images,” IEEE Access, vol. 12, no. February, pp. 16386–16396, 2024, doi: 10.1109/ACCESS.2024.3359235.

N. Network, “Automatic Identification of Down Syndrome Using Facial Images with Deep Convolutional Neural Network,” Diagnostics, vol. 10, no. 7, p. 487, 2020, doi: https://doi.org/10.3390/diagnostics10070487.

K. Rezaee, “Machine learning and facial recognition for down syndrome detection : A comprehensive review,” Comput. Hum. Behav. Reports, vol. 17, no. September 2024, p. 100600, 2025, doi: 10.1016/j.chbr.2025.100600.

“Detection of Down Syndrome Computer Vision Project.” [Online]. Available: https://universe.roboflow.com/projects-qbm9c/detection-of-down-syndrome

Dherya Bengani and Prof. Vasudha Bah, “Face Detection using Viola Jones with Haar Cascade,” Int. J. Mod. Trends Sci. Technol., vol. 6, no. 11, pp. 131–134, 2020.

M. K. Kumar et al., “A Hybrid Model for Face Detection Using HAAR Cascade Classifier and Single Shot Multi-Box Detectors Based on Open CV,” Int. Res. J. Multidiscip. Scope, vol. 5, no. 1, pp. 650–660, 2024, doi: 10.47857/irjms.2024.v05i01.0304.

A. H. Alyousef, “Implementing Face Detector using Viola-Jones Method,” SSRG Int. J. Electr. Electron. Eng., vol. 10, no. 7, pp. 140–147, 2023, doi: 10.14445/23488379/IJEEE-V10I7P113.

N. A. Mohd Ariffin, U. A. Gimba, and A. Musa, “Face Detection based on Haar Cascade and Convolution Neural Network (CNN),” J. Adv. Res. Comput. Appl., vol. 38, no. 1, pp. 1–11, 2025, doi: 10.37934/arca.38.1.111.

A. Jauhari, D. R. Anamisa, and Y. D. P. Negara, “Detection system of facial patterns with masks in new normal based on the Viola Jones method,” J. Phys. Conf. Ser., vol. 1836, no. 1, 2021, doi: 10.1088/1742-6596/1836/1/012035.

C.-H. Choi, “Face Detection Using Haar Cascade Classifiers Based on Vertical Component Calibration,” Human-centric Comput. Inf. Sci., vol. 12, no. 11, p. 18, 2022, doi: https://doi.org/10.22967/HCIS.2022.12.011.

H. Hassan et al., “Review and classification of AI-enabled COVID-19 CT imaging models based on computer vision tasks,” Comput. Biol. Med., vol. 141, no. December 2021, p. 105123, 2022, doi: 10.1016/j.compbiomed.2021.105123.

R. Kim and E. White, “Convolutional Neural Network for Data Augmentation Convolutional Neural network for Data Augmentation,” World J. Adv. Eng. Technol. Sci., vol. 13, no. 2, pp. 870–886, 2025, doi: https://doi.org/10.30574/wjaets.2024.13.2.0528.

M. Melinda, N. A. C. Andryani, Y. Nurdin, V. Khariyunnisa, Y. Yulita, and I. K. A. Enriko, “Deep Learning Performance Analysis for Facial Expression Based Autism Spectrum Disorder Identification,” Radioelectron. Comput. Syst., vol. 4225, no. 2(110), pp. 30–40, 2024, doi: 10.32620/reks.2024.2.03.

T. Kumar, R. Brennan, A. Mileo, and M. Bendechache, “Image Data Augmentation Approaches: A Comprehensive Survey and Future Directions,” IEEE Access, vol. 12, pp. 187536–187571, 2024, doi: 10.1109/ACCESS.2024.3470122.

A. Mumuni and F. Mumuni, “Data Augmentation: A Comprehensive Survey of Modern Approaches,” Elseveir, vol. 16, 2022, doi: 10.1016/j.array.2022.100258.

F. Fahmi, M. Melinda, P. D. Purnamasari, E. Elizar, and A. Rafiki, “Recognition of EEG Features in Autism Disorder Using SWT and Fisher Linear Discriminant Analysis,” Diagnostics, vol. 15, no. 18, pp. 1–16, 2025, doi: 10.3390/diagnostics15182291.

M. Melinda, M. Oktiana, Y. Nurdin, I. Pujiati, M. Irhamsyah, and N. Basir, “Performance of ShuffleNet and VGG-19 Architectural Classification Models for Face Recognition in Autistic Children,” Int. J. Adv. Sci. Eng. Inf. Technol., vol. 13, no. 2, pp. 674–680, 2023, doi: 10.18517/ijaseit.13.2.18274.

W. Tang, J. Sun, S. Wang, and Y. Zhang, “Review of AlexNet for Medical Image Classification,” EAI Endorsed Trans. e-Learning, vol. 9, 2023, doi: 10.4108/eetel.4389.

M. S. & A. M. Monika Bansal, Munish Kumar, “Transfer learning for image classification using VGG19: Caltech-101 image data set,” J. Ambient Intell. Humaniz. Comput., vol. 14, no. 3609–3620, 2023, doi: https://doi.org/10.1007/s12652-021-03488-z.

L. Ma and Z. Long, “A Face Recognition Method Using ResNet34 and RetinaFace,” Acad. J. Comput. Inf. Sci., vol. 6, no. 10, pp. 18–23, 2023, doi: 10.25236/ajcis.2023.061003.

W. Tang, “Review of Image Classification Algorithms Based on Graph Convolutional Networks,” Remote Sens., vol. 13, no. 22, pp. 1–51, 2021, doi: https://doi.org/10.3390/rs13224712.

A. Zhou et al., “Multi-head attention-based two-stream EfficientNet for action recognition,” Multimed. Syst., vol. 29, no. 2, pp. 487–498, 2023, doi: 10.1007/s00530-022-00961-3.

S. Duhan, P. Gulia, N. S. Gill, T. A. Oliveira, and P. Kumar, “Evaluation of lightweight and efficient deep learning models for plant disease classification IN Evaluation of Lightweight and Efficient Deep Learning Models for Plant Disease Classification AR IN,” Discov. Internet Things, 2026, doi: https://doi.org/10.1007/s43926-026-00310-0.

A. Bauravindah and D. H. Fudholi, “Lightweight Models for Real-Time Steganalysis: A Comparison of MobileNet, ShuffleNet, and EfficientNet,” J. RESTI (Rekayasa Sist. dan Teknol. Informasi), vol. 5, no. 158, 2024, doi: 10.29207/resti.v8i6.6091.

F. Mehmood, S. Ahmad, and T. K. Whangbo, “An Efficient Optimization Technique for Training Deep Neural Networks,” Mathematics, vol. 11, no. 6, p. 22, 2023, doi: 10.3390/math11061360.

W. Cheng, R. Pu, and B. Wang, “AMC: Adaptive Learning Rate Adjustment Based on Model Complexity,” Mathematics, vol. 13, no. 4, 2025, doi: 10.3390/math13040650.

Y. Dong, “Practice and Optimization of Deep Learning Model Training,” Int. J. Comput. Sci. Inf. Technol., vol. 5, no. 1, pp. 48–58, 2025, doi: 10.62051/ijcsit.v5n1.05.

J. T. D. C. J. Romero, “A comprehensive survey of loss functions and metrics in deep learning,” Artif Intell Rev, vol. 58, no. 195, 2025, doi: ttps://doi.org/10.1007/s10462-025-11198-7.

A. Jentzen and A. Riekert, “A proof of convergence for stochastic gradient descent in the training of artificial neural networks with ReLU activation for constant target functions,” Zeitschrift fur Angew. Math. und Phys., vol. 73, no. 5, 2022, doi: 10.1007/s00033-022-01716-w.

D. Chicco and G. Jurman, “The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation,” BMC Genomics, vol. 21, no. 1, pp. 1–13, 2020, doi: 10.1186/s12864-019-6413-7.

B. Saif, “A Comprehensive Survey on AlexNet improvements and fusion techniques,” Fusion Pract. Appl., vol. 17, no. 2, pp. 123–146, 2025, doi: 10.54216/FPA.170210.