RESUMEN
BACKGROUND: Robot-assisted implant surgery has emerged as a novel digital technology, and the accuracy need further assessment. PURPOSE: This study aimed to compare the accuracy of single dental implant placement between a novel semi-active robot-assisted implant surgery (RAIS) method and the conventional free-hand implant surgery (FHIS) method through a multicenter, randomized controlled clinical trial. MATERIALS AND METHODS: Patients requiring single dental implant placement were recruited and randomized into RAIS and FHIS group. Deviations at the platform, apex, and angle between the planned and final implant positions were assessed in both groups. Additionally, the evaluation of instrument and surgical complications was examined. RESULTS: A total of 140 patients (median age: 35.35 ± 12.55 years; 43 males, 97 females) with 140 implants from four different research centers were included, with 70 patients (70 implants) in the RAIS group and 70 patients (70 implants) in the FHIS group. In the RAIS and FHIS groups, the median platform deviations were 0.76 ± 0.36 mm and 1.48 ± 0.93 mm, respectively (p < 0.001); median apex deviations were 0.85 ± 0.48 mm and 2.14 ± 1.25 mm, respectively (p < 0.001); and median angular deviations were 2.05 ± 1.33° and 7.36 ± 4.67°, respectively (p < 0.001). Similar significant difference also presented between RAIS and FHIS group in platform vertical/horizontal deviation, apex vertical/horizontal deviation. Additionally, implants with self-tapping characteristics exhibited significantly larger deviations compared with those without self-tapping characteristics in the RAIS group. Both RAIS and FHIS methods demonstrated comparable morbidity and safety pre- and post-operation. CONCLUSIONS: The results indicated that the RAIS method demonstrated superior accuracy in single dental implant placement compared with the FHIS method. Specifically, RAIS exhibited significantly smaller deviations in platform, apex, and angular positions, as well as platform and apex vertical/horizontal deviations. This clinical trial was not registered prior to participant recruitment and randomization. https://www.chictr.org.cn/showproj.html?proj=195045.
RESUMEN
Skin cancer has a high mortality rate, and early detection can greatly reduce patient mortality. Convolutional neural network (CNN) has been widely applied in the field of computer-aided diagnosis. To improve the ability of convolutional neural networks to accurately classify skin lesions, we propose a multiscale feature fusion model for skin lesion classification. We use a two-stream network, which are a densely connected network (DenseNet-121) and improved visual geometry group network (VGG-16). In the feature fusion module, we construct multireceptive fields to obtain multiscale pathological information and use generalized mean pooling (GeM pooling) to reduce the spatial dimensionality of lesion features. Finally, we built and tested a system with the developed skin lesion classification model. The experiments were performed on the dataset ISIC2018, which can achieve a good classification performance with a test accuracy of 91.24% and macroaverages of 95%.