Oliveira Suzana, Siegl Regina, Moreira Kelly, Calvo Ana Flávia, Imparato José Carlos Pettorossi, Gimenez Thais
Graduate Program in Pediatric Dentistry, Faculdade São Leopoldo Mandic, Campinas, Brazil.
Pediatric Dentistry, University of São Paulo, São Paulo, Brazil.
Caries Res. 2024 Dec 16:1-10. doi: 10.1159/000543110.
The aim of this in vitro study was to evaluate the accuracy of near-infrared imaging (NIRI) generated by the iTero Element 5D scanner for detecting proximal caries lesions in deciduous molars, compared with visual inspection (VI), bitewing (BW) radiography, and histological examination (reference standard).
Sound deciduous molars and those with caries lesions (ICDAS 0 to 5) on the proximal surface were included, while teeth with marginal crest breakdown, restoration on the mentioned surfaces, or extensive dentin resorption were excluded. A total of 182 deciduous molars, divided into 91 pairs of first and second molars, were evaluated by two blinded examiners. Two distinct thresholds were considered for the assessment of caries detection methods: enamel lesion (D1) and dentin lesion (D3). Specificity, sensitivity, accuracy, and area under the ROC curve (AUC) were calculated for each method in the statistical analysis.
All evaluated methods demonstrated high specificity at D1, with 1.00 for VI and 0.95 for NIRI and BW, and at D3, 0.99 for BW and 0.98 for NIRI and VI, without statistical differences. The sensitivity and accuracy of NIRI at D1 were 0.44 and 0.55, while those of VI were 0.46 and 0.58, and of BW were 0.60 and 0.68, respectively. And sensitivity and accuracy of NIRI at D3 were 0.14 and 0.69, while those of VI were 0.37 and 0.77, and of BW were 0.51 and 0.82, respectively. No difference in sensitivity and accuracy was observed between NIRI and VI at D1 (p = 0.589); however, NIRI presented the lowest accuracy at D3. At D1, no statistical difference was observed between the AUC of BW and VI (p = 0.1124), nor between NIRI and VI (p = 0.2523). However, at D3, statistical differences were observed between the AUCs of the three evaluated methods: VI versus NIRI (p = 0.0005), VI versus BW (p = 0.0281), and NIRI versus BW (p < 0.0001).
NIRI showed accuracy comparable to VI at the D1 threshold but was less effective than BW radiography at both D1 and D3 thresholds.
本体外研究的目的是评估iTero Element 5D扫描仪生成的近红外成像（NIRI）在检测乳牙近端龋损方面的准确性，并与目视检查（VI）、咬合翼片（BW）X线摄影及组织学检查（参考标准）进行比较。
纳入表面健康的乳牙及近端表面有龋损（国际龋病检测和评估系统[ICDAS]0至5级）的乳牙，排除边缘嵴破坏、上述表面有修复体或广泛牙本质吸收的牙齿。共有182颗乳牙，分为91对第一和第二乳磨牙，由两名盲法检查者进行评估。在评估龋病检测方法时考虑了两个不同的阈值：釉质病变（D1）和牙本质病变（D3）。在统计分析中计算每种方法的特异性、敏感性、准确性及ROC曲线下面积（AUC）。
所有评估方法在D1时均显示出高特异性，VI为1.00，NIRI和BW为0.95；在D3时，BW为0.99，NIRI和VI为0.98，均无统计学差异。NIRI在D1时的敏感性和准确性分别为0.44和0.55，VI分别为0.46和0.58，BW分别为0.60和0.68。NIRI在D3时的敏感性和准确性分别为0.14和0.69，VI分别为0.37和0.77，BW分别为0.51和0.82。在D1时，NIRI和VI在敏感性和准确性方面无差异（p = 0.589）；然而，NIRI在D3时准确性最低。在D1时，BW和VI的AUC之间无统计学差异（p = 0.1124），NIRI和VI之间也无差异（p = 0.2523）。然而，在D3时，三种评估方法的AUC之间存在统计学差异：VI与NIRI（p = 0.0005）、VI与BW（p = 0.0281）、NIRI与BW（p < 0.0001）。
NIRI在D1阈值时显示出与VI相当的准确性，但在D1和D3阈值时均不如BW X线摄影有效。
