Toward environmentally efficient orthodontic imaging: assessing JPEG compression effects on automated artificial intelligence and manual cephalometric analysis.

Authors

Sertaç Aksakallı, Halit Hami Öz, Erhan Demir, İrem Aydın

Published in

European journal of orthodontics. Volume 48. Issue 6. Oct 06, 2026.

Abstract

This study evaluates the impact of various JPEG compression levels (1.0, 0.8, 0.5, and 0.3) on the stability and accuracy of cephalometric analyses performed by a manual expert and three commercial AI (artificial intelligence) software packages (CranioCatch, CEPHX, and WEBCEPH).
Forty digital lateral cephalometric radiographs were compressed into four quality levels using a standardized web application (ImageMagick v7.1.0). A total of 10 measurements, categorized into skeletal, dental, and soft-tissue parameters, were analyzed. Measurement stability was assessed using the Friedman test, while agreement with the manual reference standard was evaluated using intraclass correlation coefficients (ICCs) and mean absolute error.
Stability was maintained across most parameters up to the 0.3 compression level (70% quality reduction) (P > .05). Significant discrepancies were found between AI engines relative to the manual standard. The highest variation occurred in the Nasolabial angle, with a 20.95° difference between the manual expert and WEBCEPH. Baseline agreement with the manual standard ranged from low to excellent across parameters (ICC 0.50-0.96), with the highest agreement for interincisal angle and U1/SN, and the lowest for SNA, nasolabial angle, and U1 to NA (mm). While skeletal parameters showed high inter-software consistency, soft-tissue and dental inclination assessments revealed a significant lack of standardization.
High JPEG compression (down to 0.3) produced no clinically meaningful measurement changes in high-quality radiographs, providing a 70%-80% file size reduction supporting "Green Radiology" goals. However, baseline diagnostic variations between AI platforms necessitate consistent single-software use for longitudinal tracking.

PMID:
42975185
Bibliographic data and abstract were imported from PubMed on 03 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.