Retrospective study on mandible morphology towards improving implant design

Gillingham, Ryan (2019-04)

Thesis (MEng)--Stellenbosch University, 2019.

Thesis

ENGLISH ABSTRACT: Pre-surgery planning is crucial to the success of orthognathic surgery. With the advancement in 3D imaging modalities, modern methods in predicting a pathological mandible’s ideal geometry have improved. As a result, the design of patient-specific implants has become more commonplace. Before this, standard sized implants were inevitably used. Despite these enhanced virtual reconstruction techniques, limitations in these methods still exist. The most effective approach during virtual reconstruction is to replace the pathological area with the unaffected region on the opposite half of the mandible. This mirroring method becomes futile in scenarios where the disturbance overlaps the mandibular midline. Therefore, the aim of this study was to develop a virtual mandibular reconstruction technique for the purpose of aiding surgeons during implant design, whilst accounting for this limitation. It was proposed that this could be achieved by performing a retrospective investigation on the population’s mandibular structure and developing prediction models based on statistical methods. Two prediction models were formulated: a sparse prediction model (SPM) and a statistical shape model (SSM). The SPM offers a prediction of important unknown mandibular measurements when receiving the values of known measurements as an input, whilst the SSM provides an estimate for the full mandibular geometry after receiving mandibular coordinates as an input. The effectiveness of these techniques was tested by predicting missing anatomical features on subjects not part of the dataset used to create the models. The tests took place for two scenarios: the first being for when the plane of symmetry is available and the second for when it’s not. For the first scenario of testing, the mirroring method was also implemented, where the resulting accuracy served as the baseline. For both testing scenarios, the SSM clearly outperformed the SPM. Thus, there is no clear benefit in using the SPM over the SSM for virtual reconstruction scenarios. For the first scenario of testing, the SSM compared similarly to the mirroring method, where no significant difference was found between their respective accuracies (p<0.05). The difference between these two methods lies in their restriction of use. Whilst the mirroring method is constrained to situations such as the first scenario, the SSM has no such restriction. For the second scenario, the SSM produced estimations with accuracies similar to the first scenario, thus producing consistent accuracies in geometry prediction regardless of the area being reconstructed. It was therefore concluded that a SSM of the mandible presents itself as a modular virtual reconstruction technique that successfully accounts for the limitations found in current methods.

AFRIKAANSE OPSOMMING: Beplanning voor chirurgie is noodsaaklik vir die sukses van ortognatiese chirurgie. Modern metodes om n patologiese kakebeen se ideale geometrie te skep het verbeter met die bevordering van 3D-beeldmodaliteite. Daarom het die ontwerp van pasiënt-spesifieke inplantings meer algemeen geword. Dit was voorhein onvermydelik om standaard-grootte inplantings te gebruik. Ten spyte van hierdie verbeterde virtuele rekonstruksie tegnieke is daar steeds beperkinge met die metodes. Die mees effektiewe rekonstruksie manier is om die patologiese gebied te vervang met die onaangeraakde streek op die teenoorgestelde helfte van die kakebeen. Nietemin, hierdie spieëlmetode is nutteloss in gevalle waar die kakebeen versteuring die middellyn oorvleuel. Daarom was die doel van hierdie studie om ’n virtuele kakebeen rekonstruksie tegniek te ontwikkel om inplantingsontwerp te ondersteun tydens chirurgie wat hierdie beperkings voorkom. Dit is voorgestel dat dit bereik kan word deur ’n terugwerkende ondersoek op die populasie se kakebeen struktuur en die ontwikkeling van ’n voorspellingsmodelle wat gebaseer is op statistiese metodes. Twee voorspellingsmodelle is geformuleer: ’n skars voorspellingsmodel (SVPM) en ’n statistiese vormmodel (SVM). Die SVPM bied ’n voorspelling van belangrike onbekende kakebeen metings wanneer die waardes van bekende metings as ’n inset ontvang word. Die SVM lewer ’n voorspelling van die volledige kakebeen geometrie nadat die kakebeen koördinate as inset ontvang is. Die effektiwiteit van altwee tegnieke is getoets deur ontbrekende anatomiese kenmerke te voorspel op persone wat nie deel was van die datastel wat gebruik is om die modelle te skep nie. Die toetse het plaasgevind vir twee scenario’s: die eerste is vir wanneer die simmetrievlak beskikbaar is en die tweede vir wanneer dit nie is nie. Vir die eerste scenario van toetse is die spieëlmetode ook geïmplementeer en die resultate dien as die akkuraatheid basis vir die res van die toetse. Vir beide toets scenario’s het die SVM beter presteer as die SVPM. Daarom is daar geen duidelike voordeel in die gebruik van die SVPM oor die SVM vir virtuele rekonstruksiescenario’s nie. Vir die eerste scenario van toetse het die SVM geen beduidende verskil gewys in vergelyking met die spieëlmetode (met ’n akkuraatheid van p <0.05). Die twee metodes is onderskeibaar deur hulle verskillende beperkings van gebruik. Terwyl die spieëlmetode beperk word tot gevalle soos die eerste scenario, die SVM het geen sodanige beperkings nie. Vir die tweede scenario het die SVM skattings gelewer met n soortgelyke akkuratheid as die eerste scenario, en het dus konsekwente akkuraatheid in geometriese voorspellings gelewer ongeag die gebied wat herbou word. Daar is tot die gevolgtrekking gekom dat ’n SVM van n kakebeen n suksesvolle tegniek is vir modulère virtuele rekonstruksie wat die beperkings van huidige metodes voorkom.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/106066
This item appears in the following collections: