1Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India. Email: rosciashiney@mepcoeng.ac.in
2Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India. Email: ppalniladevi@mepcoeng.ac.in
Background: Early detection of breast cancer remains a major challenge due to subtle radiological variations and limited labeled medical data. This work presents an enhanced breast cancer classification model using hybrid deep learning framework that integrates modified VGG based feature extraction with an XGBoost classifier to improve diagnostic performance.
Methods: The VGG architecture is enhanced through batch normalization, reduced filter redundancy, and dropout regularization to improve feature generalization. The modified VGG network is enhanced to capture discriminative spatial and textural patterns from mammographic images. These extracted features are then classified using XGBoost, a powerful gradient boosting algorithm, which enhances model accuracy by effectively handling complex patterns and variations in the data to improve classification robustness and mitigate overfitting.
Results: Extensive experiments on benchmark breast imaging datasets demonstrate that the proposed hybrid model outperforms conventional machine learning approaches. Performance metrics such as accuracy, F1 score, sensitivity, specificity, and AUC show consistent improvements, confirming the effectiveness of integrating modified VGG features with the XGBoost classifier.
Conclusion: The findings indicate that combining deep spatial representations with gradient boosted decision trees provides an effective strategy for early breast cancer diagnosis.
Keywords: Breast cancer, Extreme Gradient Boosting, Visual Geometry Group architecture, Contrast Limited Adaptive Histogram Equalization, Deep learning.
How to cite this article: Roscia Jeya Shiney J, Palniladevi P. A Hybrid Deep Learning Approach Integrating Modified VGG Features With XGBoost For Early Breast Cancer Diagnosis. Int J Drug Deliv Technol. 2026;16(2): 741-750. DOI: 10.25258/ijddt.16.2.79
Source of support: Nil.
Conflict of interest: None