Hybrid Deep Learning Model for The Classification of Bone Tumour
DOI:
https://doi.org/10.26437/0bpx6b49Keywords:
Bone tumour. convolutional neural networks. deep learning. morphological. vision transformerAbstract
Purpose: The purpose of this study is to propose a new hybrid deep learning model for classifying bone tumour histology.
Design/Methodology and Approach: The model uses 253 samples from both tumour and non-tumour datasets, with a ResNet-50 backbone to extract localised structural features and a Transformer head to capture global context. It then merges these features and runs them through a softmax classifier for binary tumour classification, which meets clinical needs for precise and understandable histopathological diagnoses. All images were resized to a fixed resolution of 224 × 224 pixels. After resizing, pixel values were normalised to a range of [0, 1] by dividing each pixel by 255. Further normalisation was done using the mean and standard deviation values from the ImageNet dataset: mean = [0.485, 0.456, 0.406] and standard deviation = [0.229, 0.224, 0.225].
Research Limitation: This study did not examine multi-class classification of bone tumour subtypes or leverage self-supervised pretraining to reduce reliance on labelled data.
Findings: The results suggest strong potential for clinical decision support, especially for accurately detecting malignant tissues, as demonstrated in an evaluation of a simulated dataset. The model achieved 99.0% accuracy, 100% recall, a 99.0% F1 score, and a perfect AUC of 1.00.
Practical Implication: This study provides solutions for complex bone tumour analysis, offering significant benefits to the medical industry.
Social Implications: These demands include technical progress accompanied by inclusive data governance, transparent annotation practices, equitable deployment strategies, and sustained dialogue among technologists, clinicians, patients, and policymakers.
Originality / Value: This study lays the foundation for interpretable AI systems in digital pathology, demonstrating that combining CNNs, Transformers, and morphological knowledge can deliver powerful, interpretable solutions for complex medical image analysis.
References
REFERENCES
Anisuzzaman, D. M., Rahman, M. M., & Islam, M. M. (2020). Comparative analysis of CNN architectures for osteosarcoma histopathology image classification. Computers in Biology and Medicine, 121, 103759. https://doi.org/10.1016/j.compbiomed.2020.103759 DOI: https://doi.org/10.1016/j.compbiomed.2020.103759
Borji, A., Kronreif, G., Angermayr, B., & Hatamikia, S. (2025). Advanced hybrid deep learning model for enhanced evaluation of osteosarcoma histopathology images. Frontiers in medicine, 12, 1555907. DOI: https://doi.org/10.3389/fmed.2025.1555907
Chen, X., Li, Y., & Wang, Z. (2023). Self-supervised vision transformers for interpretable morphological phenotyping in bone tumour histopathology. IEEE Transactions on Medical Imaging, 42(3), 789–799. https://doi.org/10.1109/TMI.2023.3245678
Dhopavkar, G. M., Bhoyar, D. B., Kriplani, P., Patil, A. A., Nilawar, A., Prayagi, S., ... & Bhoyar, V. D. (2025). Transforming Medical Imaging with GANs: A Study on Synthetic Data Generation for Brain Tumour Diagnosis. African Journal of Applied Research, 11(6), 1-15. DOI: https://doi.org/10.26437/w96rzt82
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., &Houlsby, N. (2021). An image is worth 16x16 words: Transformers for image recognition at scale. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 45–67. https://doi.org/10.1109/CVPR46437.2021.00975 DOI: https://doi.org/10.1109/CVPR46437.2021.00975
Gbedawo, W. V., Dzikunu, A., & Nyamadi, M. (2024). Scalability and efficiency of deep learning models on high-performance computing clusters: bibliometric analysis. African Journal of Applied Research, 10(2), 283-305. DOI: https://doi.org/10.26437/ajar.v10i2.807
Hinterwimmer, S., Müller, F., & Weber, M. (2024). Multimodal transformer framework integrating clinical metadata and radiographs for bone tumour classification. Medical Image Analysis, 85, 102752. https://doi.org/10.1016/j.media.2023.102752 DOI: https://doi.org/10.1016/j.media.2023.102752
Kansara, M., Teng, M. W., Smyth, M. J., & Thomas, D. M. (2014). Translational biology of osteosarcoma. Nature Reviews Cancer, 14(11), 722–735. https://doi.org/10.1038/nrc3838 DOI: https://doi.org/10.1038/nrc3838
Kawauchi, S., Yamamoto, Y., & Tanaka, K. (2024). Vision transformer-based classification of benign and malignant lesions in 18F-FDG PET/CT images. European Journal of Nuclear Medicine and Molecular Imaging, 51(1), 112–121. https://doi.org/10.1007/s00259-023-06345-9
Kiani Kalejahi, B., Khan, S., & Zakirov, R. (2026). Bone-CNN: A Lightweight Deep Learning Architecture for Multi-Class Classification of Primary Bone Tumours in Radiographs. Biomedicines, 14(2), 299. DOI: https://doi.org/10.3390/biomedicines14020299
Li, L., Sun, Y., Luo, J., Liu, M. (2024). Circulating immune cells and risk of osteosarcoma: A Mendelian randomization analysis. Front. Immunol, 15, 1381212. DOI: https://doi.org/10.3389/fimmu.2024.1381212
Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., ...& Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical Image Analysis, 42, 60–88. https://doi.org/10.1016/j.media.2017.07.005 DOI: https://doi.org/10.1016/j.media.2017.07.005
Park, C. W., Oh, S. J., Kim, K. S., Jang, M. C., Kim, I. S., Lee, Y. K., ... & Seo, S. W. (2022). Artificial intelligence-based classification of bone tumors in the proximal femur on plain radiographs: System development and validation. PLoS One, 17(2), e0264140 DOI: https://doi.org/10.1371/journal.pone.0264140
Saka, S., & Boddupalli, S. (2025). A Deep Learning Framework for Efficient Bone Cancer Detection and Classification. In 2025 1st International Conference on Secure IoT, Assured and Trusted Computing (SATC) (pp. 1-10). IEEE. DOI: https://doi.org/10.1109/SATC65530.2025.11137353
Sampath, K., Rajagopal, S., & Chintanpalli, A. (2024). A comparative analysis of CNN-based deep learning architectures for early diagnosis of bone cancer using CT images. Scientific Reports, 14(1), 2144. DOI: https://doi.org/10.1038/s41598-024-52719-8
Song, W., Wang, X., Guo, Y., Li, S., Xia, B., Hao, A. (2024). CenterFormer: A Novel Cluster Center Enhanced Transformer for Unconstrained Dental Plaque Segmentation. IEEE Trans. Multimed, 26, 10965–10978. DOI: https://doi.org/10.1109/TMM.2024.3428349
Wang, H., He, Y., Wan, L., Li, C., Li, Z., Li, Z., ... & Tu, C. (2025). Deep learning models in classifying primary bone tumors and bone infections based on radiographs. NPJ Precision Oncology, 9(1), 72. DOI: https://doi.org/10.1038/s41698-025-00855-3
Wang, H., Zhang, L., & Zhao, J. (2022). Incorporating morphological features for improved bone tumour classification. Computers in Biology and Medicine, 142, 105231. https://doi.org/10.1016/j.compbiomed.2021.105231
Wang, Y., Zhang, H., Li, J., & Chen, S. (2023). Hybrid CNN-ViT model for bone tumour classification from CT images. Journal of Medical Imaging and Health Informatics, 13(2), 345–356. https://doi.org/10.1007/s10916-023-01845-2
Zhou, X., Li, Y., Zhang, Q., Wang, H., & Chen, J. (2023). SEAGNET: A supervised edge-attention guidance segmentation network for accurate segmentation of bone malignant tumour lesions in MRI images. Computers in Biology and Medicine, 157, 106660. https://doi.org/10.1016/j.compbiomed.2023.106660 DOI: https://doi.org/10.1016/j.compbiomed.2023.106660
Downloads
Published
Issue
Section
License
Copyright (c) 2026 AFRICAN JOURNAL OF APPLIED RESEARCH
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
By submitting and publishing your articles in the African Journal of Applied Research, you agree to transfer the copyright of the Article from the authors to the Journal ( African Journal of Applied Research).
