AI for Tooth Detection in Cleft Lip and Palate Patients
Author Information
Author(s): Arslan Can, Yucel Nesli Ozum, Kahya Kaan, Sunal Akturk Ezgi, Germec Cakan Derya, Choi Siu Wai
Primary Institution: Yeditepe University, Istanbul, Turkey
Hypothesis
The performance of the AI-based software in detecting and numbering teeth in cleft lip and palate patients does not differ significantly from expert ground truth annotations, irrespective of age or gender.
Conclusion
The AI system was effective in detecting and numbering teeth in cleft lip and palate patients, but further refinement is required for improved accuracy in the cleft region, particularly on the left side.
Supporting Evidence
- The AI system demonstrated high overall sensitivity (0.98 ± 0.03) and precision (0.96 ± 0.04).
- No statistically significant differences were found between age groups (p > 0.05).
- Challenges were observed in the maxillary left region, which exhibited higher false positive and false negative rates.
- Unilateral left clefts were more common in the study population, affecting AI performance.
- Gender comparisons showed no significant differences in AI performance, but some trends were noted.
Takeaway
This study shows that AI can help find and count teeth in kids with cleft lip and palate, but it needs to get better at doing this in certain tricky areas.
Methodology
Analyzed 100 panoramic radiographs using AI software, comparing its performance to expert annotations.
Potential Biases
Potential gender-based differences in AI performance were noted, particularly in younger age groups.
Limitations
The AI system faced challenges in accurately identifying teeth in the maxillary left region, which had higher false positive and false negative rates.
Participant Demographics
100 patients (52 males, 48 females) aged 6 to 15 years with unilateral or bilateral cleft lip and palate.
Statistical Information
P-Value
p > 0.05 for age groups; p < 0.001 for true positive rates across age groups.
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website