
New AI technology to help doctors detect lung cancer earlier
Just as a car's diagnostic system quickly identifies technical issues, the new technology aims to assist radiologists in swiftly and accurately identifying lung diseases.
With increasing pressure on healthcare systems and a global push for the early detection of lung cancer, researchers at OUH (Odense University Hospital) and SDU (University of Southern Denmark) are collaborating to revolutionise the field. The AI-RAPTOR project, a PhD initiative led by Frederik Duedahl, combines artificial intelligence with medical expertise to optimise the screening process and support radiologists in their work.
An increasing need for screening
Lung cancer screening using low-dose CT scans is standard in countries like Canada, England, and South Korea. In Denmark, we face a challenge: adopting the same inclusion criteria as international projects would require approximately 180,000 Danes to undergo annual CT scans. This would place a massive burden on the relatively small number of thoracic radiologists in the country.
-We have only around 20 specialised thoracic radiologists in Denmark, explains Frederik Duedahl.
-It would require quadrupling this number to handle the increased volume of scans that a potential screening programme would entail.
This is the backdrop for the project, which Frederik Duedahl has paused his surgical training to focus on fully. The project and its potential continue to grow.
AI as decision support
Through the AI-RAPTOR project, Frederik Duedahl and his team aim to address the issue of too few thoracic radiologists to meet the growing number of scans by developing a model capable of identifying lung cancer.
The technology is designed to function as a decision-support tool for radiologists, alleviating their workload and ensuring faster and more accurate diagnoses.
Meet the researcher
Frederik Duedahl is a physician and a PhD student at the Department of Clinical Research and the Department of Cardiothoracic Surgery at Odense University Hospital.
Meet the researcher
Simon Lyck Bjært Sørensen is an AI consultant and research assistant at SDU Robotics, part of the Maersk McKinney Moller Institute.
What is a Thoracic Radiologist?
A radiologist is a doctor specialising in diagnosing and treating diseases through medical imaging techniques such as X-ray, CT, MRI, and ultrasound.
A thoracic radiologist is a radiologist specialising in imaging of the thorax, which includes the lungs, trachea, and heart.
AI-RAPTOR is essentially an algorithm trained to recognise signs of lung cancer by analysing a vast number of scan images.
The technical development is led by Simon Lyck Bjært Sørensen, an AI consultant and the project's lead developer. He likens the model to a car’s diagnostic system:
-Instead of dismantling the entire car to find the issue, we use the onboard computer for a clear diagnosis. Similarly, our model can quickly and effectively point to potential problems in a CT scan.
From findings to quantification
The model can not only identify tumours in the lungs but also describe its findings. For example, it can briefly specify the location and size of a tumour. This enables the generation of detailed reports for radiologists to use in their assessments.
-A single scan can produce several detailed reports. This saves time and ensures that we utilise the data optimally, explains Simon Lyck Bjært Sørensen.
A strong interdisciplinary effort
The project is a collaboration between SDU's Faculty of Health Sciences and Faculty of Engineering, proving to be a significant advantage.
-When we collaborate across disciplines, as Simon and I do in this project, we can swiftly clarify both medical and technical questions, says Frederik Duedahl.
This interdisciplinary approach enables the integration of medical insights with advanced technology, creating a robust solution.
A unique platform
Beyond the AI model, the researchers have developed a new platform capable of handling large volumes of image data in a GDPR-compliant manner. This platform allows researchers to work with patient data and train their models without compromising data security.
-The uniqueness of our solution lies not only in its ability to detect lung cancer but also in the infrastructure we’ve developed, which can be applied to other diseases and types of imaging, Simon explains.
In the future, the model may help diagnose other lung conditions such as chronic obstructive pulmonary disease (COPD), pulmonary embolism, and fibrotic diseases. These developments will be pursued through future PhD projects.
Facts About Lung Cancer and Screening in Denmark
- 60% Diagnosed Late: Approximately 60% of lung cancer cases are discovered at advanced stages, making lung cancer the leading cause of cancer-related deaths globally.
- Early Detection Saves Lives: Low-dose CT scans (LDCT) of high-risk individuals can reduce mortality by identifying lung cancer earlier.
- New Screening Programme: Health authorities are planning to introduce a national lung cancer screening programme to improve early detection.
- Increased Healthcare Demand: The programme will result in an additional 100,000–150,000 CT scans annually, potentially exacerbating the shortage of specialised thoracic radiologists.
- More Radiologists Needed: To maintain the current 24-day turnaround from scan to diagnosis, the number of thoracic radiologists would need to quadruple if a screening programme is implemented.
- AI Decision Support: The AI-RAPTOR technology serves as a decision-support tool to help radiologists analyse scans faster and more accurately. The final assessment remains with the radiologist, combining technological input with professional expertise.
The future of AI and healthcare
Supported by the Danish Health Authority's PLUS project, researchers are now testing their model against radiologists’ evaluations. Initial results are promising, and the project could become a game-changer for future screening programmes.
Frederik Duedahl views the project as an essential step towards next-generation screening programmes.
-AI will never replace radiologists' expertise, but it can become an indispensable part of a more efficient and comprehensive screening system that identifies more diseases in time.
Facts About the AI-RAPTOR Project
Project Partners:
Department of Cardiothoracic Surgery, OUH
Department of Pulmonary Medicine, OUH
Department of Radiology, OUH
Maersk Mc-Kinney Moller Institute, SDU
RIT/RIPA, Region of Southern Denmark
Supervisors:
Michael Stenger
Benjamin Rasmussen
Jes S. Lindholt
Rajeeth Savarimuthu
Christian Borberg Laursen
The AI-RAPTOR PhD project has received funding from the Independent Research Fund Denmark (DFF) and the Regional Innovation Fund in Southern Denmark.