Early Diagnosis of Mammogram Images Using Hybird Deep and Machine Learning Algorithm
Gomathi S, M. Manikandan, Kannan S. Shanmugam, Roopa Chandrika Rajappan · Traitement du signal · 2025
Automated early detection of breast cancer using Computer Aided Diagnosis (CAD) is the most important step in extending the cancer patient's lifetime.However, achieving this with the CAD model, high accuracy in mammogram image analysis remains a challenge with the numerous machine learning (ML) algorithms.Sometimes, the medical practitioner needs a second opinion with this automated result.This motivated me to find a robust and reliable disease classification model to enhance the classification performance.Deep learners (DL) are the most powerful tool in extracting complex features, such as subtle abnormalities in complex images.The machine learning algorithm will classify the images with the optimal feature set to overcome the overfitting problem, and in this way, it achieves a higher accuracy than the existing traditional and other machine learning algorithms.This paper presents a new framework that integrates AdaBoost (ML) with two popular CNNs: AlexNet and ResNet (DL).AlexNet is used for feature extraction, and ResNet approaches vanishing gradient issues with a residual learning framework.The proposed model is a hybrid DL combination integrated with features coming from the CNN, which captures inherent characteristics of mammogram images to be exploited by AdaBoost acknowledged for its strong classification power.The proposed hybrid model is trained and tested on a wide publicly available mammogram dataset and achieves a high classification rate in terms of sensitivity (90 %) and specificity (92.8 %) in testing.