Confiabilidade de dois aplicativos de celular para aquisição de modelos 3D dos tecidos moles faciais

Abstract

Facial scanning allows the acquisition of reliable 3D images of the patient´s anatomy, which has considerably increased diagnostic capacity. To evaluate the accuracy and precision of a simple, low-cost method, with no exposure to ionizing radiation for the acquisition of 3D facial image, ten adult patients were selected, 5 of each gender. All patients were submitted to cone beam computed tomography examination (gold standard) and were immediately scanned with Bellus3D Face (BF) and Bellus3D Dental (BD) applications. The facial models were superimposed, divided into 14 anatomical regions of interest (RAI) and compared with the aid of the Geomagic Qualify® and MeshValmet® software. Accuracy and precision were evaluated using Root Mean Square (RMS) values. Data comparison was performed using Wilcoxon's test for one sample and Friedman's ANOVA, which was also used to compare differences between applications. The Wilcoxon test for paired samples was used to assess differences between the Face+Neck and Full Head models. Data analysis was complemented by Bland-Altman plots. There was a statistically significant difference (p≤0,05) in both applications, for all RAI's and between all evaluated methods. For accuracy analysis, the smallest mean error in the BF application was 0,27mm±0,12mm in the nasal dorsum region (DN) (p=0,002). On the other hand, the area with the greatest discrepancy was the right gonial region (RGD), with a mean error of 3,1mm±2,18mm (p=0,002). Regarding the accuracy of the BD system, a smaller deviation of 0,33mm±0,27mm was observed in the DN region and in the left infraorbital region (RIOE), which also presented a deviation of 0,33mm±0,31mm (p=0,002). The area with the highest mean error was the left gonial region (RGE), with 2,24mm±2,56mm (p=0002). The BF and BD applications were not precise, as all measurements were different from zero (p≤0,05). For the precision of the BF application, the mean errors ranged from 0,18mm±0,1mm (p=0,002) for the DN region to deviations of 1,03mm±0,44mm (p=0,002) for the submandibular region (RSM). For the BD application, the errors ranged from 0,16mm±0,19mm (p=0,004) in the right infraorbital region (RIOD) to 1,42mm±1,29mm (p=0,002) in the soft pogonion region (RPM). Regarding the acceptable clinical relevance measure, commonly considered 2mm, only the bilateral gonial regions and the submandibular region showed deviations between 2 and 3mm for the two applications. It was concluded that although the three-dimensional evaluation methods by portable devices have shown to be promising, the evaluated applications did not present adequate accuracy and precision to obtain a reliable 3D facial model