ChatGPT-4 demonstrated near-perfect accuracy in extracting clinical variables from MRI and CT scans of pancreatic cyst patients, equalling manual chart review by radiologists.
Key Details
- 1Study involved 991 patients with 3,198 MRI and CT scans from Memorial Sloan Kettering (2010–2024).
- 2ChatGPT-4 extracted 9 critical clinical variables with accuracy rates of 92%–99%.
- 3AI matched the manual approach, considered the gold standard, for classifying high-risk variables such as cyst size and solid component presence.
- 4Efficiency and cost-effectiveness cited as major advantages of the AI approach.
- 5The study was published in the Journal of the American College of Surgeons and NIH/NCI partially funded the work.
Why It Matters
Demonstrating LLMs’ ability to accurately extract structured data from radiology imaging reports could streamline large-scale research and surveillance tasks, reducing costs and improving patient care efficiency. This proof-of-concept signals the increasing feasibility and trustworthiness of leveraging advanced AI models in radiological workflows.
Source
EurekAlert
Related News
•EurekAlert
AI-Powered OCT Enables Rapid 'Optical Biopsy' for Early Endometrial Cancer Detection
A team at Washington University has developed a catheter-based 3D OCT system with AI to quickly and noninvasively detect early endometrial cancers.
•EurekAlert
AI Clinical Reasoning in Diagnostics and Digital Fatigue in Healthcare
Recent JMIR features explore large language models in clinical diagnostics and digital fatigue among healthcare professionals.
•EurekAlert
KAIST, MIT, Microsoft Develop Efficient AI Image Upsampling for Robotics
KAIST, MIT, and Microsoft have created 'Upsample Anything,' a training-free AI method to restore high-resolution visual data from compressed images with up to 16x improved GPU memory efficiency.