Related Experiment Video
Updated: Aug 31, 2025
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.3K
Development of a computer-aided quality assurance support system for identifying hand X-ray image direction using
Mitsuru Sato1, Yohan Kondo2, Masashi Okamoto2
1Department of Radiological Technology, School of Health Sciences, Niigata University, 2-746 Asahimachi-dori, Chuo-ku, Niigata, Niigata, 951-8518, Japan. mitu-sato@clg.niigata-u.ac.jp.
Radiological Physics and Technology
|August 24, 2022
Summary
Radiographic incidents persist due to human error. A new computer-aided quality assurance support system accurately identifies hand image direction, improving safety in medical imaging.
Area of Science:
- Medical Imaging
- Radiography
- Artificial Intelligence
Background:
- Digitization has improved imaging convenience but not human error prevention in radiography.
- Radiographic incidents and accidents remain a concern, with current interpretation methods sometimes overlooking issues.
- Computer-aided quality assurance support systems are crucial for enhancing safety and preventing errors.
Purpose of the Study:
- To develop an accurate method for identifying hand image direction, a key component of a computer-aided quality assurance support system.
- To evaluate the performance of different image processing techniques for classifying hand directions in X-ray images.
- To enhance the reliability of radiographic quality assurance through automated image analysis.
Main Methods:
- A novel method using U-Net segmentation for background removal and binarization of hand X-ray images was developed.
- Three classification methods were evaluated: original images, histogram equalization images, and binarized images.
- A dataset of 14,236 hand X-ray images was used to classify four hand directions: upward, downward, rightward, and leftward.
Main Results:
- Classification accuracy rates were 89.20% for original images, 99.10% for histogram equalization images, and 99.70% for binarized images with U-Net segmentation.
- The method utilizing binarization images for background removal via U-Net segmentation achieved the highest accuracy.
- The developed system demonstrates high accuracy in identifying hand direction in clinical settings.
Conclusions:
- The computer-aided quality assurance support system effectively identifies hand image direction with high accuracy.
- The novel method using U-Net segmentation significantly improves the accuracy of hand direction classification compared to conventional methods.
- This technology offers a promising solution for reducing human error and enhancing quality assurance in medical radiography.
Keywords:
Computer-aided quality assurance support systemDeep convolutional neural networkHand radiographIncident preventionQuality assurance for medical imagesSemantic segmentation
