Lung cancer is the deadliest form of cancer and early detection is critical to improving survival. Currently, diagnosing lung nodules through computerized tomography (CT) imaging is time-consuming and requires expert interpretation, often delaying diagnosis. This project aims to develop a lightweight, automated deep learning model to accurately predict the growth and malignancy of lung nodules using CT scan data. By incorporating advanced neural networks, including graph-based transformers, the model will generate future versions of lung nodules and classify cancer subtypes. The system will be trained on datasets that include intersectional demographic data, helping to uncover disparities in outcomes based on race, sex, and age. This approach seeks to speed up diagnosis, reduce human error, and address equity gaps in lung cancer care.

This project could lead to earlier, more accurate lung cancer diagnoses, especially for high-risk populations. It also has the potential to reduce healthcare delays and improve survival through automated, equitable diagnostic tools.