Author: Abdullah Hidayat, Wazir Muhammad 👨🔬
Affiliation: Florida Atlantic University 🌍
Purpose: This study aims to predict Head and Neck cancer using an artificial neural network (ANN) through readily available basic health data. The goal is to uncover hidden patterns and predictors in the basic health data, which can improve early Head and Neck cancer diagnostics
Methods: We developed a multilayer artificial neural network (ANN), which is designed with three hidden layers containing 12, 13, and 14 neurons, respectively. We applied our model to demographic, lifestyle, and clinical variables, including smoking status, alcohol consumption, ethnicity, gender, occupation, and family cancer history and several other factors. These variables were extracted from the National Health Interview Survey (1997-2022) and the prostate, lung, Colorectal, and Ovarian cancer screening trials, adapting it to the context of Head and Neck cancer. These two datasets give a strong foundation for testing and developing our predictive model.
Results: Our ANN model established promising accuracy in predicting Head & Neck cancer and outperformed standard statistical methods. Specific performance metrics, such as the area under the receiver operating characteristics (ROC) curve (AUC), demonstrate the model’s prediction power. The AUC remains at 0.83 for both training and validation cohorts, indicating consistent performance across datasets. Additionally, the investigation explored potential predictors that may provide insights into risk factors associated with head and neck cancer.
Conclusion: Our ANN model can significantly contribute to the early detection and risk assessment of Head and Neck cancer. The successful and effective application of our ANN model to the two data sets highlights the merit of artificial intelligence in enhancing cancer diagnostics.