本研究利用多種深度學習模型和資料擴增技術，結合手寫與繪畫測驗，提出了一種新穎的評估方法來區分自閉症兒童與典型發展兒童。我們招募了50位台灣國小學童，其中30位為典型發展兒童，20位為自閉症兒童，並使用Beery-Buktenica視覺動作整合測試進行實驗。為了提升數據多樣性和模型的泛化能力，我們應用了多種資料擴增技術，並選擇了ConvNeXt_Small、ResNeSt50和EfficientNet_V2_Small模型，使用Optuna進行超參數調整。
我們還引入了Ensemble Learning方法，發現stacking方法在提高分類準確性和穩定性方面表現最佳，最終達到0.935的分類準確率。進一步分析顯示，畫圓形和空間概念在區分自閉症與典型發展兒童方面尤為有效，並透過T檢定顯示這組圖片的分類準確率p值為0.021，顯著小於0.05。
研究結果證實，繪畫特徵能有效區分這兩類兒童，並強調了VMI測驗的可靠性。同時，這一方法具備跨文化應用潛力，為未來的治療和教育措施提供了寶貴的依據。此外，我們還建立了一個包含自閉症與典型發展兒童繪圖的資料集，為未來研究提供了重要資源。本研究對自閉症兒童的早期識別和治療具有跨文化應用價值，並對標準化測驗流程的設計做出了貢獻。

This study utilizes various deep learning models and data augmentation techniques, combined with handwriting and drawing tests, to propose a novel assessment method for distinguishing children with autism from typically developing children. We recruited 50 elementary school students in Taiwan, including 30 typically developing children and 20 children with autism, and conducted experiments using the Beery-Buktenica Developmental Test of Visual-Motor Integration. To enhance data diversity and model generalization, we applied multiple data augmentation techniques and selected ConvNeXt_Small, ResNeSt50, and EfficientNet_V2_Small models, with hyperparameters tuned using Optuna.
We also introduced Ensemble Learning methods and found that the stacking method performed best in improving classification accuracy and stability, ultimately achieving a classification accuracy of 0.935. Further analysis showed that drawing circles and spatial concepts were particularly effective in distinguishing between children with autism and typically developing children, with a t-test revealing a p-value of 0.021, significantly less than 0.05.
The results confirm that drawing features can effectively distinguish these two groups, highlighting the reliability of the VMI test. This method also has cross-cultural application potential, providing valuable support for future therapeutic and educational measures. Additionally, we created a dataset containing drawings by children with autism and typically developing children, providing important resources for future research. This study offers cross-cultural application value for the early identification and treatment of children with autism and contributes to the design of standardized test procedures.

Akiba, T., Sano, S., Yanase, T., Ohta, T., & Koyama, M. (2019). Optuna: A Next-generation Hyperparameter Optimization Framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (pp. 2623-2631). https://doi.org/10.1145/3292500.3330701
American Academy of Pediatrics. (2024). Developmental milestones. Retrieved from https://www.healthychildren.org/English/ages-stages/Pages/default.aspx
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Publishing.
Apollo Behavior. (2024). Understanding fine motor skill difficulties in autism. Retrieved from https://apollobehavior.com/understanding-fine-motor-skill-difficulties-in-autism
Beery, K. E., Buktenica, N. A., & Beery, N. A. (2010). Beery-Buktenica developmental test of visual-motor integration (6th ed.). Pearson.
Brownlee, J. (2020). Impact of dataset size on deep learning model skill and performance estimates. Machine Learning Mastery. Retrieved from https://machinelearningmastery.com/impact-of-dataset-size-on-deep-learning-model-skill-and-performance-estimates/
Cai, M., Li, M., Xiong, Z., Zhao, P., Li, E., & Tang, J. (2022). An Advanced Deep Learning Framework for Video-Based Diagnosis of ASD. In: Wang, L. et al. (Eds) MICCAI 2022, LNCS 13434, pp. 434–444. Springer, Cham. https://doi.org/10.1007/978-3-031-16440-8_42
CDC. (2023). Autism Data and Statistics. Retrieved from https://www.cdc.gov/autism/data-research/index.html
Chen, S.-L., Chen, S.-H., & Tseng, S.-C. (2023). The intervention effects of the forced memory handwriting teaching method on third-grade elementary school students with writing difficulties. Journal of Special Education Research, 48(3). https://doi.org/10.6172/BSE.202311_48(3).0003
Dawson, G., Rogers, S., Munson, J., Smith, M., Winter, J., Greenson, J., et al. (2010). Randomized, controlled trial of an intervention for toddlers with autism: The Early Start Denver Model. Pediatrics, 125(1), e17–e23. https://doi.org/10.1542/peds.2009-0958
Duan, Y., Zhao, W., Luo, C., Liu, X., Jiang, H., Tang, Y., Liu, C., & Yao, D. (2022). Identifying and Predicting Autism Spectrum Disorder Based on Multi-Site Structural MRI With Machine Learning. Frontiers in Human Neuroscience, 15, Article 765517. https://doi.org/10.3389/fnhum.2021.765517
Ganaie, M. A., Hu, M., Malik, A. K., Tanveer, M., & Suganthan, P. N. (2022). Ensemble deep learning: A review. arXiv. https://arxiv.org/abs/2104.02395
Green, R. R., Bigler, E. D., Froehlich, A., Prigge, M. B. D., Travers, B. G., Cariello, A. N., … Lainhart, J. E. (2015). Beery VMI performance in autism spectrum disorder. Child Neuropsychology, 22(7), 795–817. https://doi-org.nthulib-oc.nthu.edu.tw/10.1080/09297049.2015.1056131
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778). https://doi.org/10.1109/CVPR.2016.90
Hendr, A., Ozgunalp, U., & Kaya, M. E. (2023). Diagnosis of Autism Spectrum Disorder Using Convolutional Neural Networks. Electronics, 12(3), 612. https://doi.org/10.3390/electronics12030612
Hilton, C. L., Zhang, Y., Whilte, M. R., Klohr, C. L., & Constantino, J. N. (2012). Motor impairment in sibling pairs concordant and discordant for autism spectrum disorders. Journal of Autism and Developmental Disorders, 42(8), 1798–1807. https://doi.org/10.1007/s10803-012-1623-8
Hoque, S., Fairhurst, M.C., & Razian, M.A. (2004). Modulating population granularity for improved diagnosis of developmental dyspraxia from dynamic drawing analysis. Ninth International Workshop on Frontiers in Handwriting Recognition, 26-31.
Johnson, C. P., & Myers, S. M. (2007). Identification and evaluation of children with autism spectrum disorders. Pediatrics, 120(5), 1183-1215. https://doi.org/10.1542/peds.2007-2361
Ling, K. Y., Wong, J., & Chen, A. (2024). Identifying Chinese Handwriting Characteristics for Detecting Children with Autism. In Proceedings of the 39th ACM/SIGAPP Symposium on Applied Computing (pp. 856-865). https://doi.org/10.1145/3605098.3635893
Liu, Z., Mao, H., Wu, C.-Y., Feichtenhofer, C., Darrell, T., & Xie, S. (2022). A ConvNet for the 2020s. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 11976-11986). https://doi.org/10.1109/CVPR42600.2022.01201
Lord, C., Brugha, T. S., Charman, T., Cusack, J., Dumas, G., Frazier, T., Jones, E. J. H., Jones, R. M., Pickles, A., & State, M. W. (2020). Autism spectrum disorder. Autism, 24(8), 1956-1976. https://doi.org/10.1177/1362361320971107
Mackenzie, N. (2022). Understanding and Supporting Young Writers from Birth to 8. Aussie Childcare Network.
Mayo Clinic. (2024). Autism spectrum disorder. Mayo Clinic. Retrieved from https://www.mayoclinic.org/diseases-conditions/autism-spectrum-disorder/symptoms-causes/syc-20352928
Mottron, L., Belleville, S., & Ménard, E. (2005). Local bias in autistic subjects as evidenced by graphic tasks: Perceptual hierarchization or working memory deficit? Journal of Child Psychology and Psychiatry, and Allied Disciplines, 40(7), 743–755. https://doi.org/10.1111/1469-7610.00490
PyTorch. (2021). Transforming and augmenting images. PyTorch Documentation. Retrieved from https://pytorch.org/vision/stable/transforms.html
Schneider, H. J., Reiners, C., Schneider, M., & Röhrig, B. (2010). Impact of hypothyroidism on the severity of obstructive sleep apnea: A meta-analysis. Journal of Clinical Endocrinology & Metabolism, 95(4), 1344-1352. https://doi.org/10.1210/jc.2009-2446
Shin, S., Crapse, T. B., Mayo, J. P., & Sommer, M. A. (2009). Visuomotor integration. In: Binder, M. D., Hirokawa, N., & Windhorst, U. (Eds.), Encyclopedia of Neuroscience. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-29678-2_6384
Student. (1908). The probable error of a mean. Biometrika, 6(1), 1-25. https://doi.org/10.1093/biomet/6.1.1
Tan, M., & Le, Q. (2019). EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. In Proceedings of the 36th International Conference on Machine Learning (pp. 6105-6114). https://doi.org/10.1145/3605098.3635893
Tan, M., & Le, Q. (2021). EfficientNetV2: Smaller Models and Faster Training. In Proceedings of the 38th International Conference on Machine Learning (pp. 10096-10106). https://doi.org/10.48550/arXiv.2104.00298
Toback, M. et al. (2023). Neurodevelopmental Disorders of the Cerebellum: Autism Spectrum Disorder. In: Marzban, H. (eds) Development of the Cerebellum from Molecular Aspects to Diseases. Contemporary Clinical Neuroscience. Springer, Cham. https://doi.org/10.1007/978-3-031-23104-9_21
Wang, H., & Avillach, P. (2021). Diagnostic Classification and Prognostic Prediction Using Common Genetic Variants in Autism Spectrum Disorder: Genotype-Based Deep Learning. JMIR Medical Informatics, 9(4), e24754. https://doi.org/10.2196/24754
Wilkes, B.J., Archer, D.B., Farmer, A.L. et al. Cortico-basal ganglia white matter microstructure is linked to restricted repetitive behavior in autism spectrum disorder. Molecular Autism 15, 6 (2024). https://doi.org/10.1186/s13229-023-00581-2
Xie, S., Girshick, R., Dollár, P., Tu, Z., & He, K. (2017). Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 1492-1500). https://doi.org/10.1109/CVPR.2017.634
Zhang, H., Wu, C., Zhang, Z., Zhu, Y., Zhang, Z., Lin, H., ... & Sun, J. (2020). ResNeSt: Split-attention networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 2736-2746). https://doi.org/10.1109/CVPR42600.2020.00282
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016). Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2921-2929).