Related Experiment Video
Updated: Dec 15, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
686
Deep convolutional neural network-based anomaly detection for organ classification in gastric X-ray examination
Ren Togo1, Haruna Watanabe2, Takahiro Ogawa2
1Education and Research Center for Mathematical and Data Science, Hokkaido University, N-12, W-7, Kita-ku, Sapporo, 060-0812, Japan.
Computers in Biology and Medicine
|July 14, 2020
Summary
A novel deep convolutional neural network anomaly detection model effectively classifies esophagus and stomach X-ray images. This deep autoencoding Gaussian mixture model (DAGMM) shows high accuracy for organ classification in gastric X-ray examinations.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Distinguishing between esophagus and stomach images in gastric X-ray examinations is crucial for diagnosis.
- Traditional methods may face challenges due to imbalanced datasets, with fewer esophagus images compared to stomach images.
Purpose of the Study:
- To evaluate a deep convolutional neural network (CNN)-based anomaly detection model for classifying esophagus and stomach images from gastric X-ray examinations.
- To assess the model's performance against other established anomaly detection techniques.
Main Methods:
- A deep autoencoding Gaussian mixture model (DAGMM) with a convolutional autoencoder architecture was developed for organ classification.
- The model was trained and tested on 6012 subjects' X-ray images.
- Performance was compared with original DAGMM, AnoGAN, and One-Class Support Vector Machine (OCSVM) using Inception-v3 features.
Main Results:
- The proposed CNN-based anomaly detection model achieved high performance metrics: 0.956 sensitivity, 0.980 specificity, and 0.968 harmonic mean.
- These results surpassed those of the original DAGMM (0.907 harmonic mean), AnoGAN (0.834 harmonic mean), and OCSVM (0.934 harmonic mean).
Conclusions:
- The developed deep convolutional neural network-based anomaly detection model demonstrates significant potential for clinical application in organ classification.
- The proposed method offers a robust solution for distinguishing between esophagus and stomach images, particularly in scenarios with imbalanced data.
