IADR Abstract Archives
Application of Deep Learning Technology to Support the Diagnosis of Gingivitis
Objectives: This study aimed to develop and assess a solfware which supports diagnosing and provides dental recommendation for gingivitis using machine learning.
Methods: The total sample size was 720 separated into three stags of study. In all stages, every subjects were clinically examined and taken an standard intraoral photos set for the reference diagnosis (the gold standard diagnosis). At the first stage, 508 gingivitis patients were recruited to develop the dataset for training the software. At the second stage, datas from 112 subjects were utilized to evaluate the accuracy of diagnosis provided by the software after being trained. At the last stage, we used datas from 100 subjects to assess the reliability of providing dental recommendations of the software for subjects with and without gingivitis. Statistical analyses consisted of sensitivity and specitivity of diagnoses provided by the software, and the percentage of accurate dental recommendations for each type of diagnosis (with or without gingivitis).
Results: Compared to dentist’s diagnoses and software’s diagnoses has the sensitivityof 77%, the specificity of 72,9% and the accuracy of 74,5%. According to the diagnoses, the percentage of the accurate dental recommendations given by the software was 100%
Conclusions: This Solfware could be used as an equipment to diagnose and give a dental advice for Gingivitis
Methods: The total sample size was 720 separated into three stags of study. In all stages, every subjects were clinically examined and taken an standard intraoral photos set for the reference diagnosis (the gold standard diagnosis). At the first stage, 508 gingivitis patients were recruited to develop the dataset for training the software. At the second stage, datas from 112 subjects were utilized to evaluate the accuracy of diagnosis provided by the software after being trained. At the last stage, we used datas from 100 subjects to assess the reliability of providing dental recommendations of the software for subjects with and without gingivitis. Statistical analyses consisted of sensitivity and specitivity of diagnoses provided by the software, and the percentage of accurate dental recommendations for each type of diagnosis (with or without gingivitis).
Results: Compared to dentist’s diagnoses and software’s diagnoses has the sensitivityof 77%, the specificity of 72,9% and the accuracy of 74,5%. According to the diagnoses, the percentage of the accurate dental recommendations given by the software was 100%
Conclusions: This Solfware could be used as an equipment to diagnose and give a dental advice for Gingivitis