Int J Oral Maxillofac Implants. 2023 Nov 1;0(0):1-17. doi: 10.11607/jomi.10597. Online ahead of print.
PURPOSE: The primary aim of this study is to evaluate the correspondence between an Artificial Intelligence driven new tool prediction and the clinician’s evaluation in the immediate loading suitability of curves recorded during implant insertion in an in vitro test. The secondary aim is to analyse peak insertion torque (pIT) and variable torque work (VTW) values of the implants used for the in vitro study.
MATERIAL AND METHODS: The study was performed on artificial bone blocks of solid rigid polyurethane without cortical layer with four different densities. Five types of implants with different macrogeometries were used. A total of 140 implants (7 implants of each type in the four polyurethane blocks) were inserted. Immediately after implant placement the insertion curves were classified by the operator as suitable or non-suitable for immediate loading. In a second moment the same curves were analyzed by the new AIT that classified them as belonging to YES or NO class. For each implant pIT and VTW were also recorded.
RESULTS: The correspondence between surgeon and AIT evaluation was 99,3% with only one false-negative reported by the algorithm analysis. The sensitivity resulted 98.95%, the specificity 100%, positive predictive value 100% and negative predictive value 97.8%. Mean pIT of the whole sample was 34.19 + 19.43 Ncm while mean VTW was 2266.89 + 1993.73 Ncm. Statistically significant differences were found between implant systems in the whole sample and when divided by polyurethane block density.
CONCLUSIONS: AIT showed a high level of accuracy in the prediction of immediate loading suitability of insertion curves examined. All the implants used in the in vitro test were able to reach good levels of primary stability, excluding when inserted in the less dense polyurethane block. Clinical studies conducted in larger samples and with more surgeons involved are necessary to confirm these results.