Harnessing ensemble deep learning models for precise detection of gynaecological cancers

Abstract

Problem considered

The accurate and timely identification of gynaecological cancers is critical for improving patient outcomes and increasing survival rates. However, diagnostic imaging for these conditions is complex and prone to human error, necessitating advanced computational methods to enhance diagnostic reliability.

Methods

This study proposes an ensemble framework combining two state-of-the-art deep learning models, ResNet50 and Inception V3, for robust gynaecological malignancy detection. The synergistic integration of these models aims to leverage their strengths, significantly improving diagnostic performance. The models were trained and validated on a comprehensive dataset of medical images, including histopathology slides and radiological scans. The ensemble model's performance was rigorously evaluated using key metrics, including sensitivity, specificity, and overall diagnostic accuracy.

Results

The ensemble model achieved remarkable diagnostic accuracy, with results showing 99.8 % accuracy, 99.6 % sensitivity, and 99.9 % specificity. In comparison, the individual performance of ResNet50 and Inception V3 models was substantially lower. This demonstrates the effectiveness of the ensemble approach in detecting a wide range of gynaecological cancers, including ovarian and cervical malignancies.