Large language models like XLNet show promise in automating CPT code assignment for interventional radiology procedures.
Key Details
- 1Assigning CPT codes for interventional radiology is time-consuming and prone to error, with coding taking an average of 12 minutes per procedure and up to 80% of bills containing errors.
- 2Researchers used XLNet (an open-source LLM) to predict CPT codes from post-procedure reports, focusing on embolization and catheterization cases.
- 3The study analyzed two datasets: one with 1,600 embolization reports (17 CPT codes) and one combining embolization and catheter cases (5,600 reports; 42 CPT codes).
- 4XLNet performed best on vascular embolization and occlusion (code 37243) and central venous-access procedures (code 36597), with some codes being more challenging.
- 5The open-source and local deployment aspects of XLNet support data privacy and accessibility for institutions of all sizes.
Why It Matters
Automating CPT coding in interventional radiology using AI could reduce administrative burden, minimize errors, improve billing accuracy, and facilitate secure, institution-specific solutions. Such advancements may streamline workflows and improve financial outcomes in radiology departments.
Source
Radiology Business
Related News
•AuntMinnie
Study: Computer Vision Models Best LLMs in Chest CT Breast Abnormality Detection
Computer vision models (CVMs) surpass large language models (LLMs) in accurately labeling incidental breast abnormalities on chest CT scans.
•Cardiovascular Business
Radiology Maintains Lead in FDA-Cleared AI Algorithms, Cardiology Follows
Radiology remains the top specialty for FDA-cleared AI, with cardiology as a strong second, particularly in cardiovascular imaging.
•AuntMinnie
Deep Learning Models Rival Radiologists for Pancreatic Cancer Detection on CT
Deep-learning models achieved comparable or superior accuracy to experienced radiologists in detecting pancreatic cancer on CT scans, especially for small tumors.