Breast Cancer Detection Using Ultrasound Images

Abstract

This study presents an automated 3D breast cancer detection framework utilizing transrectal ultrasound imaging. The proposed approach combines multi-atlas registration with statistical texture priors for accurate segmentation. The atlas database includes annotated breast images from previous cases with segmented breast surfaces. Texture features are extracted using orthogonal Gabor filter banks to enhance the robustness of feature detection. Stage-specific tumor features are utilized to train a hybrid CNN-ResNet model, which ensures precise detection and segmentation of breast tumours in new patient images. Superpixel segmentation is then applied to refine tumor boundaries, enabling detailed and accurate tumor localization. The proposed method provides an efficient and reliable tool for early breast cancer detection, aiming to support improved diagnostic outcomes and clinical decision-making.

Introduction

Breast cancer remains a major global health concern, and early detection is crucial for improving survival rates. Traditional imaging methods—mammography, MRI, and especially ultrasound—are widely used, but challenges like image artifacts, operator dependence, and misinterpretation often limit their accuracy.

With advances in AI and deep learning, modern diagnostic systems now provide improved precision. Techniques such as CNNs, ResNet, DenseNet, EfficientNet, hybrid models, and attention-based frameworks have demonstrated high accuracy in classifying breast tumors. Several studies report accuracy levels above 94–96%, highlighting the growing role of AI in medical imaging.

The proposed research introduces a multi-stage automated 3D breast cancer detection framework using ultrasound images. It integrates:

• Gabor texture filters for robust feature extraction

• A hybrid CNN–ResNet architecture for tumor classification

• Superpixel-based segmentation to improve tumor boundary detection

A detailed methodology includes preprocessing, augmentation, dataset balancing, MobileNet-based feature extraction, and model training using SGD. The model achieved:

• Test accuracy: 94.9%

• Precision: 97.04%

• Recall: 97.62%

• ROC-AUC: 98.2%

Furthermore, multi-atlas registration and 3D visualization improved tumor localization with high segmentation metrics (DSC: 92.4%, Jaccard: 89.7%). The system also reduced processing time by 25%, making it efficient for clinical use.

The literature review highlights several major contributions in AI-based breast cancer detection, showing the effectiveness of deep networks (ResNet, EfficientNet, DenseNet) and hybrid models. Overall, this research demonstrates that integrating AI with medical imaging can significantly improve diagnostic reliability, assist radiologists, and support earlier, more accurate breast cancer detection.

Conclusion

This study presents an efficient and scalable deep learning-based approach for automated breast cancer detection. By leveraging the pre-trained MobileNet model for feature extraction and employing techniques such as data augmentation and bottleneck caching, the proposed methodology achieves high classification accuracy while maintaining computational efficiency. The experimental results demonstrate the model\'s robustness, with an overall test accuracy of 94.9%, a precision of 97.04%, and an ROC-AUC score of 98.2%, highlighting its effectiveness in distinguishing between benign and malignant cases. Furthermore, the deployment of the trained model as a web-based application using Flask and Gunicorn enables real-time breast cancer diagnosis, offering a user-friendly interface for seamless image uploads and classification. This solution reduces dependency on manual histopathological analysis and facilitates early detection, potentially improving patient outcomes. Future work will focus on enhancing the model\'s interpretability, incorporating additional imaging modalities, and refining misclassification handling to further improve diagnostic accuracy.

References

[1] L. Shen et al., \"End-to-End Training for Whole Mammogram Classification Using Convolutional Neural Networks,\" IEEE Transactions on Medical Imaging, vol. 39, no. 5, pp. 1291–1300, 2020. [2] J. Arevalo et al., \"Representation Learning for Mammogram Classification With Convolutional Neural Networks,\" IEEE Transactions on Biomedical Engineering, vol. 64, no. 12, pp. 2600–2611, 2017. [3] S. Hussain et al., \"Breast Cancer Detection Using Hybrid Deep Learning Model,\" IEEE Access, vol. 8, pp. 224712–224725, 2020. [4] R. Jamthikar et al., \"Automated Detection of Breast Cancer Using Machine Learning Techniques,\" IEEE Access, vol. 9, pp. 87256–87270, 2021. [5] T. Wu et al., \"Breast Cancer Detection Using Deep Residual Networks,\" IEEE Journal of Biomedical and Health Informatics, vol. 25, no. 3, pp. 911–919, 2021. [6] M. Al-Dhabyani et al., \"Deep Learning for Mammographic Breast Cancer Diagnosis Using VGG16,\" IEEE Access, vol. 8, pp. 151908–151920, 2020. [7] Y. Liu et al., \"Breast Cancer Detection Using Deep Learning with Class Activation Maps,\" IEEE Transactions on Medical Imaging, vol. 39, no. 11, pp. 3327–3336, 2020.. [8] A. Ahmad et al., \"A Hybrid Deep Learning Model for Breast Cancer Detection,\" IEEE Access, vol. 8, pp. 103191–103200, 2020 [9] H. Xu et al., \"Breast Cancer Detection Using Multi-Scale CNNs,\" IEEE Transactions on Image Processing, vol. 30, pp. 4863–4874, 2021. [10] M. Shahanas and R. Sudharson, \"Deep Transfer Learning for Mammographic Breast Cancer Detection,\" IEEE Access, vol. 9, pp. 176198–176209, 2021. [11] H. Chen et al., \"Deep Neural Networks for Breast Cancer Diagnosis Using Ultrasound Images,\" IEEE Transactions on Biomedical Engineering, vol. 68, no. 3, pp. 937–947, 2021. [12] L. Yang et al., \"Hybrid Attention Mechanisms for Breast Cancer Classification,\" IEEE Transactions on Medical Imaging, vol. 40, no. 11, pp. 3205–3216, 2021. [13] J. Qiu et al., \"Breast Cancer Detection Using Capsule Networks with Limited Data,\" IEEE Access, vol. 9, pp. 51312–51321, 2021. [14] W. Zhang et al., \"Breast Cancer Diagnosis Using Dual-Input CNN with Texture and Spatial Features,\" IEEE Access, vol. 9, pp. 75634–75643, 2021. [15] F. Zhou et al., \"EfficientNet for Breast Cancer Detection in Mammograms,\" IEEE Transactions on Medical Imaging, vol. 40, no. 12, pp. 3362–3372, 2021. [16] R. Sun et al., \"Deep Learning-Based Multi-View Mammogram Analysis for Breast Cancer Detection,\" IEEE Transactions on Biomedical Engineering, vol. 68, no. 5, pp. 1340–1351, 2021 [17] A. Das et al., \"Attention-Guided CNNs for Breast Cancer Detection in Mammograms,\" IEEE Access, vol. 9, pp. 68212–68225, 2021. [18] T. Wang et al., \"Hybrid CNN-RNN Model for Breast Cancer Detection,\" IEEE Transactions on Medical Imaging, vol. 40, no. 10, pp. 3043–3054, 2021 [19] P. Gupta et al., \"Explainable AI for Breast Cancer Detection Using Grad-CAM,\" IEEE Access, vol. 9, pp. 70327–70335, 2021. [20] Y. Jin et al., \"Deep Learning for Early Detection of Breast Cancer Using Thermal Imaging,\" IEEE Transactions on Biomedical Engineering, vol. 68, no. 7, pp. 1848–1857, 2021.

Copyright

Copyright © 2025 Dr. Priyadarshan Dhabe , Atharv Asalkar, Prathamesh Bansode, Niraj Baviskar, Atharva Belote, Hamlin Nadar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.