Research
Seeing More in Every Scan: AI for Earlier Detection of Lung Cancer and More
Seeing More in Every Scan: AI for Earlier Detection of Lung Cancer and More
Lead Researcher: Ge Wang
Institution: Rensselaer Polytechnic Institute
The Challenge
Lung cancer is the leading cause of cancer-related death worldwide, even though screening can significantly reduce mortality. Today’s screening programs face major barriers — including high false-positive rates, limited participation, and a shortage of radiologists.
At the same time, CT scans contain far more information than is typically used. In addition to detecting lung cancer, they can reveal early signs of heart disease and other conditions. But most existing AI tools are narrow, focusing on a single task and relying only on imaging — leaving valuable clinical data unused.
The Approach
In collaboration with RPI Prof. Pingkun Yan and others, Professor Wang’s team is developing an AI system that analyzes CT scans alongside clinical data — such as age, smoking history, and medical records — to deliver a more complete picture of patient health.
The system performs multiple chest CT related tasks at once, including detecting lung nodules, predicting cancer risk, and assessing cardiovascular disease. By learning from large-scale, real-world data, it can identify subtle patterns that may not be clearly visible to clinicians.
How Empire AI Makes This Possible
Empire AI enables the team to train larger, more advanced models using high-resolution 3D imaging and extensive clinical datasets.
- Train models with 500M–1B+ parameters
- Analyze 100,000+ CT scans with linked clinical records
- Run multitask models across 17+ screening objectives
- Reduce experiment time from weeks to days
This scale allows researchers to move faster and build more accurate, generalizable systems.
Potential Impact
This work could significantly improve how diseases are detected and managed:
- Earlier and more accurate detection of lung cancer
- Better identification of cardiovascular risk
- Fewer false positives in screening programs
- More efficient workflows for healthcare providers
Over time, it may enable more effective, data-driven screening at population scale — improving outcomes while reducing costs.