Integrating Regression Models and Deep Learning for Lung Cancer Risk Prediction

Early diagnosis is vital for successful treatment of lung cancer, as the disease remains one of the most common causes of death by cancer worldwide. Historically, the prediction of lung cancer has been based on regression models and disemboweled by the fact that it cannot optimally handle complex data. In contrast, deep learning techniques have exhibited great potential in developing precise predictive models of disease based on more complicated forms of data; nevertheless, these methods have been less rigorously applied specifically for predicting lung cancer risks. This project aims to combine conventional regression models and deep learning techniques to achieve better performance with more generalizable accuracy for lung cancer risk prediction. A hybrid framework will be developed, exploiting the strengths of regression models in being interpretable and deep learning for its superior prediction. Our proposed methodology will also tackle the issue of high-dimensional imaging data needed for lung cancer risk prediction. These advances in notably conjunctional regression models and deep learning may contribute to better prediction models for lung cancer risk by enhancing the accuracy and translation ability of the developed scores. It will help improve the identification of people at high risk for lung cancer and, on that note, also target the population likely to benefit, thereby ensuring early controls with better outcomes.