Comparative Study of Methods of Feature Extraction from Keypoints for the Classification of General Movements in Newborns

• André Valente de Cristo UFAM
• Eulanda Miranda dos Santos UFAM
• Rafael Giusti UFAM
• Ayrles Silva Gonçalves Barbosa Mendonça UFAM

DOI: https://doi.org/10.5753/eniac.2025.13795

Abstract

General movements assessment (GMA) is a widely used method for predicting motor dysfunction, particularly cerebral palsy. In recent years, several studies have made progress in automating GMA by using manually extracted features from key points identified in videos combined with either shallow or deep machine learning models. However, there is still no consensus in the literature regarding the best feature extraction method, and there is a lack of studies comparing different methods, especially for newborns. In this context, this work aims to compare three methods of feature extraction from keypoints found in the literature for the classification of general movements in newborns.

References

Doroniewicz, I., Ledwoń, D. J., Affanasowicz, A., Kieszczyńska, K., Latos, D., Matyja, M., Mitas, A. W., and Myśliwiec, A. (2020). Writhing movement detection in newborns on the second and third day of life using pose-based feature machine learning classification. Sensors, 20(21).

Einspieler, C., Peharz, R., and Marschik, P. B. (2016). Fidgety movements – tiny in appearance, but huge in impact. Jornal de Pediatria, 92(3, Supplement 1):S64–S70.

Einspieler, C. and Prechtl, H. F. R. (2005). Prechtl’s assessment of general movements: A diagnostic tool for the functional assessment of the young nervous system. Mental Retardation and Developmental Disabilities Research Reviews, 11(1):61–67.

Gholamiangonabadi, D., Kiselov, N., and Grolinger, K. (2020). Deep neural networks for human activity recognition with wearable sensors: Leave-one-subject-out cross-validation for model selection. IEEE Access, 8:133985–133994.

Hesse, N., Bodensteiner, C., Arens, M., Hofmann, U. G., Weinberger, R., and Schroeder, A. S. (2018). Computer vision for medical infant motion analysis: State of the art and rgb-d data set. In ECCV Workshops.

Huang, X., Huang, C., Yin, W., Huang, H., Xie, Z., Huang, Y., Chen, M., Fan, X., Shang, X., Peng, Z., Wan, Y., Han, T., and Yi, M. (2024). Automatic quantitative intelligent assessment of neonatal general movements with video tracking. Displays, 82:102658.

Jahn, L., Flügge, S., Zhang, D., Poustka, L., Bölte, S., Wörgötter, F., Marschik, P. B., and Kulvicius, T. (2024). Comparison of marker-less 2d image-based methods for infant pose estimation.

Ji, S., Ma, D., Pan, L., Wang, W., Peng, X., Amos, J. T., Ingabire, H. N., Li, M., Wang, Y., Yao, D., and Ren, P. (2024). Automated prediction of infant cognitive development risk by video: A pilot study. IEEE Journal of Biomedical and Health Informatics, 28(2):690–701.

Leo, M., Bernava, G. M., Carcagnı̀, P., and Distante, C. (2022). Video-based automatic baby motion analysis for early neurological disorder diagnosis: State of the art and future directions. Sensors, 22(3).

Letzkus, L., Pulido, J., Adeyemo, A., Baek, S., and Zanelli, S. (2024). Machine learning approaches to evaluate infants’ general movements in the writhing stage—a pilot study. Scientific Reports, 14.

McCay, K. D., Ho, E. S. L., Shum, H. P. H., Fehringer, G., Marcroft, C., and Embleton, N. D. (2020). Abnormal infant movements classification with deep learning on pose-based features. IEEE Access, 8:51582–51592.

McCay, K. D., Hu, P., Shum, H. P. H., Woo, W. L., Marcroft, C., Embleton, N. D., Munteanu, A., and Ho, E. S. L. (2022). A pose-based feature fusion and classification framework for the early prediction of cerebral palsy in infants. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 30:8–19.

Passmore, E., Kwong, A. L., Greenstein, S., Olsen, J. E., Eeles, A. L., Cheong, J. L. Y., Spittle, A. J., and Ball, G. (2024). Automated identification of abnormal infant movements from smart phone videos. PLOS Digital Health, 3(2):1–21.

Xia, L., Chen, C.-C., and Aggarwal, J. K. (2012). View invariant human action recognition using histograms of 3d joints. In 2012 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, pages 20–27.