Estimation of healthy bone shape and density distribution from partial inputs for implant design

Date
2022-04
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH SUMMARY: When reconstructing segmental bone loss, segmentation and surface reconstruction require extensive specialist knowledge to be repeated for each new patient. This has proven to be time-consuming and cost-ine cient throughout literature and practice. Statistical modelling is widely used in biomedical elds for automated segmentation and is a viable alternative for reconstructing healthy bone anatomy in the absence of healthy contralateral geometry. Therefore, as part of this study, statistical models of shape and appearance were constructed from sample data based on femur and tibia data of the male and female South African population, and their application in automated segmentation, reconstruction and density estimation was investigated. The study uses a novel combination of an active shape and a mean appearance model to estimate missing bone geometry and density distribution from sparse inputs simulating segmental bone loss around the diaphyseal area. Estimations of diaphyseal resections were obtained by probabilistic tting of the active shape model to sparse inputs consisting of proximal and distal bone data on computed tomography images. The resulting shape estimates of the diaphyseal resections were then used to map the mean appearance model to the patients' missing bone geometry, constructing density estimations. The models constructed reproduced the shape and density distribution of the population with an average error below 1.47 mm and a 90 % density t. Resected bone surfaces were estimated with an average error below 1.64 mm, and density distributions were approximated above 84 % of the intensity of the original target images. These results fall within the acceptable tolerance limits of reconstructive surgery and appear promising for practical use in patient-speci c implant design.
AFRIKAANS OPSOMMING: Tydens die herstel van segmentele beenverlies, vereis segmentering en oppervlak herkonstruksie uitgebreide spesialiskennis wat vir elke nuwe pasiënt herhaal moet word. Dit is bekend as 'n tydrowend en koste-ondoeltre ende aktiwiteite in literatuur sowel as in praktyk. Statistiese modellering word algemeen gebruik in biomediese velde vir geoutomatiseerde segmentering en is 'n lewensvatbare alternatief vir die herkonstruksie van gesonde beenanatomie in die afwesigheid van gesonde kontralaterale geometrie. Daarom, as deel van hierdie studie, is statistiese modelle van vorm en voorkoms geskep uit steekproefdata gebaseer op femur en tibia data van die manlike en vroulike Suid-Afrikaanse bevolking en hul toepassing in geoutomatiseerde segmentering, rekonstruksie en digtheidsskatting is ondersoek. Hierdie studie gebruik 'n nuwe kombinasie van 'n aktiewe vorm en 'n gemiddelde voorkomsmodel om ontbrekende beengeometrie en digtheidsverspreiding te skat vanaf gedeeltelike insette wat segmentele beenverlies rondom die dia seale area simuleer. Beramings van dia se-reseksies is verkry deur waarskynlike passing van die aktiewe vorm model op gedeeltelike insette wat bestaan uit proksimale en distale been data op rekenaar tomogra e beelde. Die gevolglike vormskattings van die dia se-reseksies is dan gebruik om die gemiddelde voorkomsmodel na die pasiënte se ontbrekende beengeometrie oor te dra en digtheidsskattings te konstrueer. Die modelle wat gekonstrueer is het die vorm en digtheidsverspreiding van die populasie weergegee met 'n gemiddelde fout onder 1.47 mm en 'n 90 % digtheidspassing. Gedeeltelike beenoppervlaktes is geskat met 'n gemiddelde fout onder 1.64 mm en digtheidsverspreidings was akkuraat tot meer as 84 % van die intensiteit van die oorspronklike teikenbeelde. Hierdie resultate val binne die aanvaarbare toleransiegrense van herkonstruktiewe chirurgie en lyk belowend vir praktiese gebruik in pasiënt-spesi eke inplantaat ontwerp.
Description
Thesis (MEng)--Stellenbosch University, 2022.
Keywords
Segmental bone repair, Automated segmentation, Statistical models, Long bone reconstruction, Healthy Bone Shape, UCTD
Citation