Scholarlink

<p><b>Objective</b> To compare the accuracy and speed of artificial intelligent (AI) cephalometric analysis with automatic landmark identification, to computer-based and paper tracing on patients diagnosed with a cleft lip and palate.</p><p><b>Materials and methods</b> In total, 39 cephalograms of patients from the cleft clinic with a repaired unilateral or bilateral cleft lip and palate were included, where 30 of the patients had a severe skeletal discrepancy. The AI software used was WebCeph. One orthodontist carried out cephalometric analysis via four methods: 1) paper; 2) computer-based; 3) AI fully automated; and 4) AI followed by manual adjustment of the landmarks as required. Each method had intra-rater reliability testing. Inter-group comparisons were performed using ANOVA followed by a post-hoc Tukey test.</p><p><b>Results</b> The landmarks most commonly requiring adjustment following automatic identification were nasion, A-point, anterior nasal spine, and upper and lower incisors. Four of the 16 cephalometric values had statistically significant differences between groups: s-n-a (<i>p</i> &lt;0.01), Ar-Go-Me (<i>p</i> &lt;0.05), S-NPNS-ANS (<i>p</i> &lt;0.05), and ANS-Me/N-Me (<i>p</i> &lt;0.01). The greatest differences occurred between AI fully automated and either paper or computer-based however. AI with manual as required was comparable to computer-based and paper. The AI methods, with or without adjustment, were both significantly quicker than computer based or paper (<i>p</i> &lt;0.01).</p><p><b>Conclusion</b> Landmark identification in WebCeph cannot be wholly relied upon in patients with repaired cleft lip and palate and significant skeletal discrepancies in comparison to paper and computer-based. However, manual adjustment of the automatically identified landmarks by a clinician provides similar results to paper and computer-based with much improved speed.</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

The accuracy and speed of artificial intelligent cephalometric software compared to computer and paper tracing in patients with cleft lip and palate

  • 期刊论文

  • 发表日期：

    2025年11月7日

查看全文
  • Ian Murphy,
  • Nigel Taylor

摘要

Objective To compare the accuracy and speed of artificial intelligent (AI) cephalometric analysis with automatic landmark identification, to computer-based and paper tracing on patients diagnosed with a cleft lip and palate.

Materials and methods In total, 39 cephalograms of patients from the cleft clinic with a repaired unilateral or bilateral cleft lip and palate were included, where 30 of the patients had a severe skeletal discrepancy. The AI software used was WebCeph. One orthodontist carried out cephalometric analysis via four methods: 1) paper; 2) computer-based; 3) AI fully automated; and 4) AI followed by manual adjustment of the landmarks as required. Each method had intra-rater reliability testing. Inter-group comparisons were performed using ANOVA followed by a post-hoc Tukey test.

Results The landmarks most commonly requiring adjustment following automatic identification were nasion, A-point, anterior nasal spine, and upper and lower incisors. Four of the 16 cephalometric values had statistically significant differences between groups: s-n-a (p <0.01), Ar-Go-Me (p <0.05), S-NPNS-ANS (p <0.05), and ANS-Me/N-Me (p <0.01). The greatest differences occurred between AI fully automated and either paper or computer-based however. AI with manual as required was comparable to computer-based and paper. The AI methods, with or without adjustment, were both significantly quicker than computer based or paper (p <0.01).

Conclusion Landmark identification in WebCeph cannot be wholly relied upon in patients with repaired cleft lip and palate and significant skeletal discrepancies in comparison to paper and computer-based. However, manual adjustment of the automatically identified landmarks by a clinician provides similar results to paper and computer-based with much improved speed.