Deep Learning Framework for Skeletal Age Classification from Pelvic Radiographs using K-fold Cross Validation and Stacking of CNN Models
Keywords:
Bone age, Deep learning, Ensemble, Fine-tuning, K-fold cross-validation, Pelvic X-raysAbstract
Recent technological developments in deep learning environments have improved bone age evaluation, making it easier and more exact than classic methods in forensic radiology. Deep convolutional neural networks are highly effective at detecting bone age; however, their complexity arises from the number of parameters they need, making them resource-intensive to run on CPUs. To address this, the proposed work utilizes the transfer learning approach to build a two-stage deep learning model based on pelvic radiographs, comprising a vital bone extraction model and an age assessment model. Initially, UNet model combined with Attention Gate extracted the pelvic girdle bones by filtering insignificant regions from pelvis X-rays. For age assessment, a smaller classifier network was first developed and evaluated using K-fold cross-validation. Subsequently, the two deep networks were built by layering the new ones to the existing framework. To enhance performance further, the outputs of both the classifiers were stacked using a dense layer called an aggregator. This meta-learner combined the strength of each model to make decisions on final prediction. The whole framework was validated to analyse its ability to categorize human age in the range of 0–19 years using the collected pelvic radiographs and achieved an average classification accuracy of 97.50%, precision of 98.25%, recall of 96.65%, and F1-score of 97.20%. Thus, the proposed framework can increase the accuracy of multi-classification tasks while leveraging the limited computational resources.
Downloads
References
M. Miranker, “A Comparison of Different Age Estimation Methods of the Adult Pelvis,” Journal of Forensic Sciences, vol. 61, no. 5, pp. 1173–1179, Jun. 2016.
B. Gümüş, E. Karavaş and O. Taydaş, “Can forensic radiological skeletal age estimation be performed by examining ischiopubic-ilioischial-iliopubic synchondrosis in computed tomography images?” PLoS ONE, vol. 17, no. 4, p. e0266682, Apr. 2022.
E. Stan, A. Enache, C.O. Muresan, V. Ciocan, S. Ungureanu, A.C. Motofelea, A. Voicu and D. Costachescu, “Age Estimation through Hounsfield Unit Analysis of Pelvic Bone in the Romanian Population,” Diagnostics, vol. 14, no. 18, pp. 2103–2103, Sep. 2024.
S.R. Vempalli, V. Meshram, R.S. Shekhawat, B. Sureka, R. Shedge, M.A. NJ, P. Setia and T. Kanchan, “3D CT based age estimation from the pubic symphyseal surface in an Indian population using the Chen et al. method,” Anthropologischer Anzeiger, vol. 81, no. 3, pp. 315–325, Jun. 2024.
V. Warrier, R. Shedge, P.K. Garg, S.G. Dixit, K. Krishan and T. Kanchan, “Machine learning and regression analysis for age estimation from the iliac crest based on computed tomographic explorations in an Indian population,” Medicine, Science and the Law/Medicine, science and the law, vol. 64, no. 3, pp. 204–216, Sep. 2023.
F. Fan, X. Dong, X. Wu, R. Li, X. Dai, K. Zhang, F. Huang and Z. Deng, “An Evaluation of Statistical Models for Age Estimation and the Assessment of the 18-year Threshold using Conventional Pelvic Radiographs,” Forensic Science International, vol. 314, pp. 110350–110350, Jun. 2020.
K. Zhang, X. Dong, F. Fan and Z. Deng, “Age estimation based on pelvic ossification using regression models from conventional radiography,” International Journal of Legal Medicine, vol. 130, no. 4, pp. 1143–1148, May 2016.
Y. Li, Z. Huang, X. Dong, W. Liang, H. Xue, L. Zhang, Y. Zhang and Z. Deng, “Forensic age estimation for pelvic X-ray images using deep learning,” European Radiology, vol. 29, no. 5, pp. 2322–2329, Nov. 2018.
R. Akhade, A. Dhanorkar, J. Chawhan and J. Khanapuri, “Bone Age Estimation System Using Deep Learning,” in Proc. of the 5th International Conference on Advances in Science and Technology (ICAST), IEEE, 2022, pp. 1-5.
L.Q. Peng, Y.C. Guo, L. Wan, T.A Liu, P. Wang, H. Zhao and Y.H. Wang, “Forensic bone age estimation of adolescent pelvis X-rays based on two-stage convolutional neural network,” International Journal of Legal Medicine, vol. 136, no. 3, pp. 797–810, Jan. 2022.
T. Bradshaw, Z. Huemann, J. Hu and A. Rahmim, “A Guide to Cross-Validation for Artificial Intelligence in Medical Imaging,” Radiology, vol. 5, no. 4, p. e220232, Jul. 2023.
S. Prusty, S. Patnaik and S. K. Dash, “SKCV: Stratified K-fold cross-validation on ML classifiers for predicting cervical cancer,” Frontiers in Nanotechnology, vol. 4, p. 972421, Aug. 2022.
T.R. Mahesh, O. Geman, M. Margala, and M. Guduri, “The stratified K-folds cross-validation and class-balancing methods with high-performance ensemble classifiers for breast cancer classification,” Healthcare Analytics, vol. 4, p. 100247, Dec. 2023.
J. Kaliappan, A.R. Bagepalli, S. Almal, R. Mishra, Y.-C. Hu and K. Srinivasan, “Impact of Cross-Validation on Machine Learning Models for Early Detection of Intrauterine Fetal Demise,” Diagnostics, vol. 13, no. 10, pp. 1692–1692, May 2023.
M.N. Raihen and S. Akter, “Comparative Assessment of Several Effective Machine Learning Classification Methods for Maternal Health Risk,” Computational Journal of Mathematical and Statistical Sciences, vol. 3, no. 1, pp. 161–176, Apr. 2024.
D. Rastogi, P. Johri, V. Tiwari and A. A. Elngar, “Multi-class classification of brain tumour magnetic resonance images using multi-branch network with inception block and five-fold cross validation deep learning framework,” Biomedical Signal Processing and Control, vol. 88, p. 105602, Feb. 2024.
N. Nivedita and S. Solanki, “Enhancing the accuracy of automatic bone age estimation using optimized CNN model on X-Ray images,” in Communications in computer and information science, 2024, pp. 329–340.
A. Davila, J. Colan and Y. Hasegawa, “Comparison of fine-tuning strategies for transfer learning in medical image classification,” Image and vision computing, vol. 146, pp. 105012–105012, Apr. 2024.
O. Ronneberger, P. Fischer and T. Brox, “U-NET: Convolutional Networks for Biomedical Image Segmentation,” in Lecture notes in computer science, 2015, pp. 234–241.
O. Oktay, J. Schlemper, L.L. Folgoc, M. Lee, M. Heinrich, K. Misawa, K. Mori, S. McDonagh, N.Y. Hammerla, B. Kainz and B. Glocker, “Attention u-net: Learning where to look for the pancreas,” 2018, arXiv:1804.03999.
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
All papers should be submitted electronically. All submitted manuscripts must be original work that is not under submission at another journal or under consideration for publication in another form, such as a monograph or chapter of a book. Authors of submitted papers are obligated not to submit their paper for publication elsewhere until an editorial decision is rendered on their submission. Further, authors of accepted papers are prohibited from publishing the results in other publications that appear before the paper is published in the Journal unless they receive approval for doing so from the Editor-In-Chief.
IJISAE open access articles are licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. This license lets the audience to give appropriate credit, provide a link to the license, and indicate if changes were made and if they remix, transform, or build upon the material, they must distribute contributions under the same license as the original.
