Related Experiment Video
Updated: May 5, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Impact of image processing techniques on deep learning-based classification accuracy of cervical vertebral
Naoki Yamada1, Yukihiro Iida2, Wakako Tome3
1Department of Orthodontics, School of Dentistry, Asahi University, 1851, Mizuho, Gifu, 501-0296, Japan. yamada530@dent.asahi-u.ac.jp.
Oral Radiology
|January 4, 2026
Summary
Image processing significantly impacts deep learning for cervical vertebral maturation (CVM) assessment. Dual-color labeling achieved the highest accuracy, improving skeletal maturity classification reliability.
Area of Science:
- Dentistry and Orthodontics
- Artificial Intelligence in Medical Imaging
Background:
- Cervical vertebral maturation (CVM) assessment is crucial for orthodontic diagnosis.
- Automated CVM staging using deep learning models requires robust image data.
Purpose of the Study:
- To evaluate the influence of various image processing techniques on deep learning model performance for CVM staging.
- To compare classification accuracies across different image types and labeling methods.
Main Methods:
- A dataset of 799 cephalometric radiographs was analyzed.
- Images were processed into standard, low-density, standard labeled, low-density labeled, and dual-color labeled variants.
- A convolutional neural network (AlexNet architecture) classified images into six CVM stages.
Main Results:
- Standard and low-density images yielded low classification accuracies (46.8% and 48.7%).
- Labeled images showed significantly improved performance: standard labeling (82.3%), low-density labeling (81.0%), and dual-color labeling (88.0%).
Conclusions:
- Dual-color labeling demonstrated the highest classification accuracy for CVM staging.
- Enhanced image labeling techniques substantially improve the reliability of automated skeletal maturity assessment using deep learning.
