Lung Cancer Longitudinal AI with Rule-Based Interpretability (CLARITY)

Work is in progress

This project will establish an open-access clinical benchmark and Virtual Imaging Trials (VIT) platform for lung cancer diagnosis (Aim 1). Furthermore, this project will develop a foundational Lung Cancer Longitudinal AI model with Rule-Based Interpretability (CLARITY) for lung cancer screening, leveraging clinical and digital patient cohorts generated from 4D human models (Aim 2). These digital cohorts will simulate diverse lung nodule characteristics (size, margin, type, and location ), each guided by real clinical findings and distributions to ensure clinical relevance and accuracy. The model will incorporate simulated longitudinal data to learn how nodules evolve over time, enabling dynamic risk predictions. Furthermore, rule-based interpretability layers aligned with clinical guidelines (Lung-RADS) will provide transparent outputs, while an actionable recommendation system will guide clinicians on follow-up reporting and care.

Objective: Develop an AI model that incorporates longitudinal data for dynamic risk predictions, coupled with rule-based interpretability to ensure actionable outputs.

Methods:

• A) Generate digital patient cohorts from 3D human phantom models to simulate longitudinal disease progression with diverse nodule types guided by real clinical findings utilizing statistical and generative models.
• B) Train the model to predict nodule growth, malignancy risk, and future changes over time.
• C) Embed rule-based decision trees (aligned with Lung-RADS) for interpretable and clinically relevant outputs. D) Develop a module to provide reporting and follow-up recommendations based on predicted risk levels.

Outcome: A dynamic and transparent AI model capable of longitudinal analysis and providing actionable insights for clinicians.

Impact Statement: The CLARITY project addresses key challenges in AI for lung cancer screening by integrating longitudinal data and rule-based interpretability into diagnostic workflows. Leveraging clinical and simulated digital patient cohorts, it aims to enhance diagnostic precision and provide actionable, guideline-aligned outputs. By fostering trust and improving clinical usability, the project aims to enhance AI-driven early detection and follow-up care, contributing to improvements in lung cancer screening and patient outcomes.