Boosting Cervical Cancer Prediction Leveraging a Hybrid FT-Transformer Model

Cervical Cancer (CC) remains a significant threat to women's health, despite being largely preventable, and is a leading cause of mortality worldwide. Early and accurate prediction is crucial for timely treatment and improved survival rates, particularly given the diverse risk factors contributing to CC development. To address this critical clinical need, we propose an innovative Hybrid FT-Transformer model that synergistically integrates a Feature Tokenization (FT) Transformer with Depthwise convolutional neural networks and Long Short-Term Memory (LSTM) networks for precise CC prediction. Explainable Artificial Intelligence (XAI) tools, including Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), provide insights into the model's decision-making process, aiding in validation by clinicians. Trained on a UCI repository dataset focusing on Hinselmann, Biopsy, and Schiller outcomes, our proposed model achieved outstanding accuracies: 99.09% for Hinselmann, 98.90% for Biopsy, and 98.75% for Schiller, using 10-fold cross-validation. These results highlight the model's superiority over current state-of-the-art approaches, with significant potential to enhance CC screening and early diagnosis, ultimately improving patient outcomes.