Proximal caries detection on smartphone display

Abstract

The aim of this study was to compare diagnostic accuracy in proximal caries detection between bitewing radiographs exported from PACS software and taken with a smartphone viewed in a smartphone display. A total of 200 proximal surfaces from digital bitewing radiographs were included in this study. Images of all radiographs were captured from a medical-grade and a common display by an iPhone 8 Plus and stored as JPEG files. Exported DICOM files were converted into JPEG format and transferred to the smartphone used for image capturing. Each proximal surface was rated by 7 observers with 5-point-scale. Weighted kappa test was used to determine intra- and inter-observer agreements. Three certified oral radiologists evaluated the same images on the medical-grade display. Obtained consensus was used to calculate sensitivity, specificity, accuracy, positive predictive value, negative predictive value and generate ROC curves. T-test and one-way ANOVA were used to compare mean AUC between dentinal and enamel caries and among three image acquiring methods.
 
 The result showed that inter- and intra-observer agreement ranged from “moderate” to “almost perfect”. Comparison of mean AUC showed significant higher value in group of exported images. While there was no significant difference between group of images captured from a medical-grade display and images captured from a common display. Significant differences between mean AUC in detection of dentinal caries were seen in all image groups. For enamel caries, only mean AUC in group of exported images was significantly higher.
 
 Detection of proximal caries should be done using directly exported images from PACS software. Captured images should be evaluated with caution since considerable factors can affect image quality.