Integration of Deep Learning Techniques with Hu Moment Invariants for Automated Cancer Stage Determination
DOI:
https://doi.org/10.59746/ena4vn48الكلمات المفتاحية:
Breast cancer، Deep learning، Mammography، Diagnosis، benign، malignant، BI-RADالملخص
: Accurate determination of breast cancer staging from mammographic imagery remains a clinically demanding challenge, owing to substantial inter-lesion variability in morphological appearance, inconsistencies in image acquisition quality, and the inherent limitations of conventional deep learning architectures in capturing diagnostically relevant structural characteristics. Predominant existing methodologies disproportionately exploit texture and intensity-based representations while systematically neglecting rotation- and scale-invariant shape descriptors that carry considerable discriminative value in tumor characterization and malignancy grading. This study seeks to develop a robust, automated breast cancer staging framework that synergistically integrates deep semantic representations with invariant morphological descriptors, with the dual objective of enhancing multi-class classification performance and enabling a more holistic and clinically interpretable characterization of mammographic lesions across BI-RADS categories. A hybrid computer-aided diagnosis (CAD) framework is proposed and evaluated on mammographic images drawn from the King Abdulaziz University (KAU) benchmark dataset. The pipeline is initialized with an autoencoder-based preprocessing stage that performs noise attenuation and perceptual image reconstruction, thereby ensuring high-fidelity input representations for downstream processing. Subsequently, high-level semantic features are extracted via the pretrained EfficientNet-B7 convolutional architecture through transfer learning, while complementary morphological representations are encoded by computing the seven Hu Moment invariant descriptors to capture affine-invariant shape properties of breast lesions. The resultant feature sets are integrated through a principled feature-level fusion strategy, yielding a unified discriminative representation for multi-stage BI-RADS classification. Comprehensive model assessment is conducted employing five standardized performance metrics: Accuracy, Precision, Recall, F1-score, and Specificity. Empirical evaluation on the KAU dataset demonstrated that the proposed hybrid framework attained an overall classification accuracy of 96.0%, accompanied by a precision of 95.5%, a recall of 97.3%, an F1-score of 96.3%, and a specificity of 99.4%. These results consistently surpass those achieved by standalone deep learning architectures and conventional handcrafted-feature methodologies, substantiating the complementary nature of fusing semantic deep representations with invariant morphological descriptors for improved lesion discrimination. The proposed hybrid CAD framework effectively harnesses the representational strengths of EfficientNet-B7 and the geometric invariance of Hu Moment descriptors to advance automated breast cancer stage determination from mammographic images. The empirical findings underscore the diagnostic value of incorporating morphological shape information into deep learning-driven classification pipelines and affirm the clinical translational potential of the proposed methodology as a decision-support instrument for radiologists engaged in BI-RADS-based breast cancer screening and diagnosis.
المراجع
[1] R. L. Siegel, K. D. Miller, and A. Jemal, “Cancer statistics, 2023,” CA: A Cancer Journal for Clinicians, vol. 73, no. 1, pp. 17–48, Jan./Feb. 2023, doi: 10.3322/caac.21763.
[2] A. Dibden, J. Offman, S. W. Duffy, and R. Gabe, “Worldwide review and meta-analysis of cohort studies measuring the effect of mammography screening programmes on incidence-based breast cancer mortality,” Cancers, vol. 12, no. 4, p. 976, Apr. 2020, doi: 10.3390/cancers12040976.
[3] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 770–778, doi: 10.1109/CVPR.2016.90.
[4] G. Huang, Z. Liu, L. van der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 4700–4708, doi: 10.1109/CVPR.2017.243.
[5] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Adv. Neural Inf. Process. Syst. (NeurIPS), 2012.
[6] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in Proc. Int. Conf. Learn. Representations (ICLR), 2015, doi: 10.48550/arXiv.1409.1556.
[7] M. Tan and Q. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” in Proc. Int. Conf. Mach. Learn. (ICML), 2019, doi: 10.48550/arXiv.1905.11946.
[8] A. Ribli et al., “Detecting and classifying lesions in mammograms with deep learning,” Scientific Reports, vol. 8, no. 1, 2018, doi: 10.1038/s41598-018-22437-z.
[9] [N. Shen, J. Wu, and S. Zhang, “Deep learning in medical image analysis: A survey,” Annu. Rev. Biomed. Eng., vol. 19, pp. 221–248, 2017, doi: 10.1146/annurev-bioeng-071516-044442.
[10] C. D. Lehman et al., “Imaging management of palpable breast abnormalities,” Amer. J. Roentgenol., vol. 196, no. 2, pp. 267–273, 2011, doi: 10.2214/AJR.10.4657.
[11] M.-K. Hu, “Visual pattern recognition by moment invariants,” IRE Trans. Inf. Theory, vol. 8, no. 2, pp. 179–187, 1962, doi: 10.1109/TIT.1962.1057692.
[12] Y. Sahiner et al., “Classification of mass and normal breast tissue: A convolution neural network classifier with spatial domain and texture images,” IEEE Trans. Med. Imaging, vol. 15, no. 5, pp. 598–610, 1996, doi: 10.1109/42.538937.
[13] [G. Litjens et al., “A survey on deep learning in medical image analysis,” Med. Image Anal., vol. 42, pp. 60–88, 2017, doi: 10.1016/j.media.2017.07.005.
[14] J. Arevalo et al., “Combining handcrafted features and convolutional neural networks for breast cancer classification,” Med. Image Anal., vol. 26, no. 1, pp. 286–301, 2016, doi: 10.1016/j.media.2015.10.001.
[15] S. Han et al., “Breast cancer classification using deep learning and hybrid features,” Comput. Biol. Med., vol. 103, pp. 40–48, 2018, doi: 10.1016/j.compbiomed.2018.10.006.
[16] R. K. Sabhara, C. P. Lee, and K. M. Lim, “Comparative study of Hu moments and Zernike moments in object recognition,” Smart Computing Review, vol. 3, no. 3, pp. 166–173, 2013, doi: 10.6029/smartcr.2013.03.003.
[17] S. A. Boudraa et al., “Improving mass discrimination in mammogram CAD system using texture information and super-resolution reconstruction,” Evolving Systems, 2020, doi: 10.1007/s12530-019-09293-0.
[18] T. Kooi et al., “Large scale deep learning for computer aided detection of mammographic lesions,” Med. Image Anal., vol. 35, pp. 303–312, 2017, doi: 10.1016/j.media.2016.11.007.
[19] H. Jung et al., “Detection of masses in mammograms using a one-stage object detector based on a deep convolutional neural network,” PLoS ONE, vol. 13, no. 9, 2018, doi: 10.1371/journal.pone.0203355.
[20] A. Saber, M. Sakr, O. M. Abo-Seida, A. Keshk, and H. Chen, “A novel deep-learning model for automatic detection and classification of breast cancer using the transfer-learning technique,” IEEE Access, vol. 9, pp. 71194–71209, 2021, doi: 10.1109/ACCESS.2021.3078880.
[21] B. Q. Huynh, H. Li, and M. L. Giger, “Digital mammographic tumor classification using transfer learning from deep convolutional neural networks,” J. Med. Imaging, vol. 3, no. 3, p. 034501, 2016, doi: 10.1117/1.JMI.3.3.034501.
[22] A. Loddo, C. Di Ruberto, and M. Kocher, “Recent advances of malaria parasites detection systems based on mathematical morphology,” Sensors, vol. 18, no. 2, p. 513, 2018, doi: 10.3390/s18020513.
[23] T. Wang et al., “A high-performance deep neural network model for BI-RADS classification of screening mammography,” Sensors, vol. 22, no. 3, p. 1160, 2022, doi: 10.3390/s22031160.
[24] S. Tran, H. T. X. Nguyen, C. Phan, H. H. Pham, and H. Q. Nguyen, “A novel transparency strategy-based data augmentation approach for BI-RADS classification of mammograms,” arXiv preprint arXiv:2203.10609, 2022, doi: 10.48550/arXiv.2203.10609.
[25] M. Gunawardhana and N. Zolek, “Performance of machine learning classification in mammography images using BI-RADS,” arXiv preprint arXiv:2311.08493, Nov. 2023, doi: 10.48550/arXiv.2311.08493.
[26] W.-C. Shen, R.-F. Chang, and W. K. Moon, “Computer-aided classification system for breast ultrasound based on Breast Imaging Reporting and Data System (BI-RADS),” Ultrasound in Medicine & Biology, vol. 33, no. 11, pp. 1688–1698, Nov. 2007, doi: 10.1016/j.ultrasmedbio.2007.05.016.
[27] F. Manigrasso et al., “Mammography classification with multi-view deep learning techniques: Investigating graph and transformer-based architectures,” Medical Image Analysis, vol. 99, p. 103320, 2025, doi: 10.1016/j.media.2024.103320.
[28] “King Abdulaziz University mammogram dataset,” Kaggle. [Online]. Available: https://www.kaggle.com/datasets/asmaasaad/king-abdulaziz-university-mammogram-dataset
