Scholarlink

Purpose <p>Lateral photographs are routinely evaluated as part of orthodontic diagnostics and treatment planning. Anthropometric measurements based on soft tissue landmarks are taken to evaluate facial features and attractiveness. The detection of these landmarks is a&#xa0;process performed by specialists and prone to intra- and inter-annotator variability. The aims of this investigation were (1)&#xa0;to train an artificial intelligence (AI) algorithm to automatically detect landmarks on lateral photographs, (2)&#xa0;to establish a&#xa0;high-quality gold standard dataset to evaluate landmark detection accuracy, and (3)&#xa0;to compare the performance of AI with that of clinical experts.</p> Methods <p>The AI algorithm was trained on a&#xa0;dataset of 991 photographs, with three clinical experts annotating 14&#xa0;soft&#xa0;tissue landmarks on each photograph. Eleven experts annotated a&#xa0;separate dataset of 56&#xa0;photographs, to establish the gold standard. Metric scaling of the photographs was achieved by transferring scaling from corresponding lateral cephalograms. Based on the detected landmarks, 11&#xa0;anthropometric measurements were taken, and the performance of the experts and AI was compared against the gold standard by comparing errors from the ground truth using Mann–Whitney&#xa0;U tests.</p> Results <p>At a&#xa0;2.0 mm threshold, the AI model achieved successful detection rates exceeding 95% for 12&#xa0;of 14&#xa0;landmarks. Compared with individual expert annotations, AI predictions showed reduced variability and lower mean radial errors for landmarks with high inter-annotator disagreement. Anthropometric measurements derived from AI predictions demonstrated smaller absolute errors than expert-based measurements.</p> Conclusion <p>This study demonstrates that AI-based landmark detection on lateral photographs can achieve accuracy comparable to expert annotations, demonstrating greater consistency for those landmarks exhibiting high inter-annotator variability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated landmark detection on lateral photographs using artificial intelligence: diagnostic accuracy compared to expert annotations

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月23日

查看全文
  • Noah Frieder Nordblom,
  • Felix Kunz,
  • Angelika Stellzig-Eisenhauer

摘要

Purpose

Lateral photographs are routinely evaluated as part of orthodontic diagnostics and treatment planning. Anthropometric measurements based on soft tissue landmarks are taken to evaluate facial features and attractiveness. The detection of these landmarks is a process performed by specialists and prone to intra- and inter-annotator variability. The aims of this investigation were (1) to train an artificial intelligence (AI) algorithm to automatically detect landmarks on lateral photographs, (2) to establish a high-quality gold standard dataset to evaluate landmark detection accuracy, and (3) to compare the performance of AI with that of clinical experts.

Methods

The AI algorithm was trained on a dataset of 991 photographs, with three clinical experts annotating 14 soft tissue landmarks on each photograph. Eleven experts annotated a separate dataset of 56 photographs, to establish the gold standard. Metric scaling of the photographs was achieved by transferring scaling from corresponding lateral cephalograms. Based on the detected landmarks, 11 anthropometric measurements were taken, and the performance of the experts and AI was compared against the gold standard by comparing errors from the ground truth using Mann–Whitney U tests.

Results

At a 2.0 mm threshold, the AI model achieved successful detection rates exceeding 95% for 12 of 14 landmarks. Compared with individual expert annotations, AI predictions showed reduced variability and lower mean radial errors for landmarks with high inter-annotator disagreement. Anthropometric measurements derived from AI predictions demonstrated smaller absolute errors than expert-based measurements.

Conclusion

This study demonstrates that AI-based landmark detection on lateral photographs can achieve accuracy comparable to expert annotations, demonstrating greater consistency for those landmarks exhibiting high inter-annotator variability.