Ensemble deep learning framework for accurate and robust cervical cancer classification

Abstract

Early and accurate cervical cancer diagnosis (CCD) is essential for improving treatment outcomes and reducing mortality. However, manual examination of cervical cytology images is labor-intensive, time-consuming, and subject to inter-observer variability. To address these limitations, this paper proposes an ensemble deep learning framework for automated cervical cancer classification. The proposed methodology integrates multiple convolutional neural network (CNN) models to exploit their complementary feature extraction capabilities, thereby improving classification accuracy and robustness. Image preprocessing and data augmentation techniques are employed to enhance image quality, reduce noise, and increase the diversity of the training dataset, enabling better model generalization. The outputs of the individual CNN models are combined through an ensemble decision strategy to generate more reliable diagnostic predictions than those obtained from a single classifier. Experimental results demonstrate that the proposed framework consistently outperforms conventional deep learning methods in terms of accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). The ensemble approach effectively reduces classification errors while maintaining stable performance across different cervical cell categories. These findings indicate that the proposed framework can serve as a reliable computer-aided diagnostic tool for cervical cancer screening, supporting clinicians in early diagnosis and improving the efficiency, consistency, and reliability of clinical decision making.