Abstract
This paper presents experimental results obtained from using weakly tuned deep learning models for skeletal age estimation from hand radiographs. By leveraging transfer learning, deep learning models were initialised with the ImageNet dataset weights and then tuned for 5 epochs on the target RSNA Bone Age dataset. Thereafter, the deep learned models were used to estimate skeletal ages and obtained a MAE of 9.96 months. With the exploration of weighted model ensemble the MAE reduced to 9.32 months. In the final ensemble, the best results achieved took into consideration predictions from one of the poor performing models. This improvement indicates that weakly tuned deep learned models still have a good prediction accuracy. Moreover, poor performing models such as the InceptionResNetV2 which achieved a MAE of 16.75 months was still needed in the ensemble for best results. This indicates that even poor performing models can still extract useful information which can be taken advantage of in ensemble situations such as weighted ensemble.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Budka, M., Gabrys, B.: Density-preserving sampling: robust and efficient alternative to cross-validation for error estimation. IEEE Trans. Neural Netw. Learn. Sys. 24(1), 22–34 (2013)
Castillo, J.C., Tong, Y., Zhao, J., Zhu, F.: RSNA Bone-age Detection using Transfer Learning and Attention Mapping, pp. 5
Iglovikov, V., Rakhlin, A., Kalinin, A., Shvets, A.: Pediatric Bone Age Assessment Using Deep Convolutional Neural Networks, vol. 11045, pp. 300–308 (2018). arXiv:1712.05053 [cs]
Kattukaran, M.S., Abraham, A.: Bone age assessment using deep. Learning 9(2), 4 (2018)
Pietka, E., Gertych, A., Pospiech, S., Cao, F., Huang, H.K., Gilsanz, V.: Computer-assisted bone age assessment: image preprocessing and epiphyseal/metaphyseal ROI extraction. IEEE Trans. Med. Imaging 20(8), 715–729 (2001)
Poznanski, Andrew K., Garn, Stanley M., Nagy, Jerrold M., Gall, John C.: Metacarpophalangeal pattern profiles in the evaluation of skeletal malformations. Radiology 104(1), 1–11 (1972)
Rakhlin, A.: Diabetic Retinopathy detection through integration of Deep Learning classification framework. bioRxiv, June 2018
Rakhlin, A., Shvets, A., Iglovikov, V., Kalinin, A.: Deep Convolutional Neural Networks for Breast Cancer Histology Image Analysis. bioRxiv, April 2018
Souza, D., Oliveira, M.M.: End-to-end bone age assessment with residual learning. In: 2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), pp. 197–203, Parana, October 2018. IEEE
Tiulpin, A., Thevenot, J., Rahtu, E., Lehenkari, P., Saarakkala, S.: Simo: automatic knee osteoarthritis diagnosis from plain radiographs: a deep learning-based approach. Scientific Report 8(1), 1–10 (2018)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Hirasen, D., Pillay, V., Viriri, S., Gwetu, M. (2020). Skeletal Age Estimation from Hand Radiographs Using Transfer Learning. In: B. R., P., Thenkanidiyoor, V., Prasath, R., Vanga, O. (eds) Mining Intelligence and Knowledge Exploration. MIKE 2019. Lecture Notes in Computer Science(), vol 11987. Springer, Cham. https://doi.org/10.1007/978-3-030-66187-8_16
Download citation
DOI: https://doi.org/10.1007/978-3-030-66187-8_16
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-66186-1
Online ISBN: 978-3-030-66187-8
eBook Packages: Computer ScienceComputer Science (R0)