2 min read
In a major advancement for cancer care, scientists at Northwestern Medicine have developed a powerful new artificial intelligence tool that can identify lung tumors with remarkable accuracy, even spotting areas that some doctors might overlook.
The tool is called “iSeg.” It’s a 3D deep learning model. Additionally, it’s trained to detect and outline lung tumors on CT scans. What makes it unique is its ability to track tumors as they move with each breath. This is vital for accurate radiation therapy.
Unlike traditional AI models that analyze still images, iSeg adapts to real-time physiological movement, enabling more precise radiation planning. Around 50% of cancer patients in the U.S. undergo radiation treatment, where accuracy can mean the difference between successful therapy and damaging healthy tissue.
Trained with Real-World Data Across Hospitals
The AI was trained using hundreds of CT scans and tumor outlines from patients treated at nine hospitals in the Northwestern Medicine and Cleveland Clinic networks. This expansive training set gives iSeg a robust edge over previous models trained on small, single-center datasets.
In testing, iSeg was challenged with new patient scans, and its results were compared to expert-drawn tumor contours. It consistently matched or outperformed doctors, and crucially, it identified additional tumor areas linked to worse outcomes if untreated.
“Accurate tumor targeting is the foundation of safe and effective radiation therapy,” said Dr. Mohamed Abazeed, the project’s senior author and chair of radiation oncology at Northwestern.
By flagging potential high-risk regions missed during manual tumor mapping, iSeg could dramatically improve cancer treatment outcomes.
“Our AI tool can help reduce delays, ensure fairness across hospitals, and potentially identify areas that doctors might miss,” said Sagnik Sarkar, lead author and senior research technologist at Feinberg.
Clinical Testing Underway
The research team is now testing iSeg in real-time clinical settings. They’re also adding new features like user feedback.
Additionally, the team plans to expand iSeg to detect tumors in the liver, brain, and prostate. They aim to use it with other imaging methods like MRI and PET scans.
“This is a foundational tool that could standardize and enhance how tumors are targeted in radiation oncology,” said Troy Teo, co-author and instructor at Feinberg.
With more testing and improvements, iSeg could be used in clinics within a few years. The goal is to close gaps in cancer care and provide more consistent, life-saving treatment.
