Radiologist You-Cheng Lin and his team at Taichung Veterans General Hospital have developed an innovative "X-ray Error Detection and Warning System." This technology uses AI deep learning models to compare the inputted imaging region with the actual positioning before X-ray exposure. This system significantly reduces human error, ensuring that each X-ray targets the correct area, thereby preventing unnecessary radiation exposure. The team collected approximately 50,000 positioning images using a camera attached to the X-ray machine and carefully selected 22,000 images for deep learning training, accounting for region-specific and data balance issues. During 43,495 applications of the system, 463 warnings were triggered, with only 0.66% (288 instances) being false positives, achieving an impressive accuracy rate of 99.3%. This technology not only enhances the accuracy of X-ray imaging but also provides a more comprehensive guarantee for patient safety and healthcare quality.

Enhancing Patient Safety with Smarter Medical Environments

This system offers three key features:

1. Accurate Image Capture and Recognition: The system employs high-resolution cameras to capture positioning images, which are transmitted via Wi-Fi to the comparison system. An image capture card extracts images from the X-ray machine's control console, performing optical character recognition (OCR) to identify the imaging region selected by the radiologist. The error detection system is activated when the X-ray room's lead door closes, ensuring that image capture occurs at the correct moment.
2. Intelligent Comparison and Classification: The team-developed comparison system matches the X-ray machine's set imaging region with the AI-recognized positioning image. The system distinguishes between non-directional and directional errors, accurately identifying region errors, left-right errors, or a combination of both.
3. Real-Time Alerts and Feedback: When an error is detected, the system immediately issues a visual and auditory warning, prompting the radiologist to verify and correct the positioning. If the comparison is correct, a "PASS" icon is displayed. The system also logs error data for subsequent review and system optimization.

A Milestone in Medical Innovation: The X-ray Error Detection and Warning System

You-Cheng Lin highlighted that while the radiation risk from diagnostic X-rays is relatively low, it can still lead to medical risks, such as incorrect diagnoses resulting in inappropriate treatment plans. As X-ray imaging continues to grow rapidly in the medical field, the increased likelihood of incorrect positioning has become a patient safety concern. The team aims to use this system to create a smarter medical environment that enhances patient safety.

Lin emphasized that the system's primary function is to detect and alert for errors without interfering with the X-ray equipment. Final judgment still rests with the radiologist. This innovative technology offers benefits to all stakeholders involved in X-ray imaging and holds significant potential for commercialization and application across various medical institutions. It represents a step forward in creating a safer and smarter healthcare environment for the future.

Resource (mandarin): AI協助偵錯預警 精準照X光防非必要輻射暴露！