In the field of radiation therapy, precision is crucial. Oncologists need to accurately map tumors before administering high-dose radiation to target cancer cells while preserving healthy tissue. This process, known as tumor segmentation, is traditionally done manually and can vary between doctors, potentially leading to critical areas being missed.
A team of scientists at Northwestern Medicine has developed an AI tool named iSeg that matches doctors in outlining lung tumors on CT scans and identifies areas some doctors might overlook. A recent study highlights iSeg's ability to segment tumors as they move with each breath, a first for 3D deep learning tools in this context.
“We’re one step closer to cancer treatments that are even more precise than any of us imagined just a decade ago,” said Dr. Mohamed Abazeed, senior author and chair of radiation oncology at Northwestern University Feinberg School of Medicine. He emphasized the goal of providing better tools for doctors.
The study was published in npj Precision Oncology on June 30. It details how iSeg was trained using CT scans from over 1,000 lung cancer patients across multiple clinics within the Northwestern Medicine and Cleveland Clinic health systems. The AI was tested on new patient scans, consistently matching expert outlines and flagging additional areas linked to worse outcomes if untreated.
“Accurate tumor targeting is the foundation of safe and effective radiation therapy,” Abazeed stated. First author Sagnik Sarkar added that automating tumor contouring could reduce delays and improve patient care by ensuring consistency across hospitals.
The research team is now testing iSeg in clinical settings and plans to expand its use to other tumor types like liver, brain, and prostate cancers. They also aim to adapt it for other imaging methods such as MRI and PET scans.
Co-author Troy Teo mentioned that clinical deployment could be possible within a couple of years, envisioning iSeg as a tool that standardizes tumor targeting in radiation oncology.
This study is titled "Deep learning for automated, motion-resolved tumor segmentation in radiotherapy."