A Meta-Learner-Integrated Stacking Voting Ensemble Network for Cervical Malignancy Classification
Kanchan Vishalkumar Wankhade, Mayuresh Gulame, Priya Khune, Aarti Paresh Pimpalkar, Sneha Singha, Mohini Kumbhar · 2024
The aggressiveness and death rate of cervical cancer pose a serious threat to a woman's health. By identifying and treating the affected tissues at the initial phases of the syndrome, a complete recovery is possible. The Papanicolaou (Pap) test is a conventional method of examining cervix tissues in order to screen for cervical cancer. Many networks for automated cervical cancer diagnosis have recently been constructed by researchers; however, the large size and poor accuracy of these single models precludes their practical implementation. Our proposal to tackle this problem involves utilizing several Inception networks as base learners and integrating their outcomes by a voting ensemble. This technique is called Voting-Stacking collective approach. The experimental results show its potential to reduce screening burden and assist pathologists in detecting diseases because they outperform the state-of-the-art technologies now in use. Furthermore, a multi-level Voting-Stacking ensemble framework is intended to enhance outcome even more. Utilizing a publically accessible dataset, our model demonstrated accuracy, precision, recall, and FI are 98.30%, 99.30%, 98.49% and 99.21 % correspondingly. The outcomes of the judgments demonstrate that the suggested framework performs admirably on pap-stained cytology pictures.