Comparative Analysis of Deep Learning Architecture with Ensemble Learning in Cranial and Mediolateral View Images
H S Sushma, Kavitha Sooda · 2023
The study explores the field of anomaly detection employing advanced methods. Mammography analysis for the identification of anomaly is examined in terms of the effectiveness of ensemble learning when combined with the powerful EfficientNet-B7 deep learning architecture. This study attempts to enhance anomaly identification accuracy by utilizing the abilities of these methods. Comparative experimental results show significant enhancements corresponding to the proposed ensemble learning model. The integrated technique outperforms the standalone EfficientNetB7 model’s 95.81% and 94.89% training and validation accuracies while attaining 98.86% and 98.29% with the proposed approach, respectively and the memory consumption has increased by approximately 18.28% with the proposed approach and 22.84% without the proposed approach. With better generalization and reduced overfitting characteristics, the ensemble model consistently shows higher accuracy over training and validation datasets. These results demonstrate the possibility of ensemble learning to improve the stability and accuracy of detection of anomalies. Selecting the right model and considering ensemble learning into account while analyzing medical images are crucial choices that could impact the diagnostic system’s reliability, effectiveness, and accuracy.