Vpliv izbire referenčne kostne strukture na natančnost registracije 3D MR modela kolčnega labruma v 3D CT model medenice : magistrsko delo

Černelič, Eva (avtor), Mekiš, Nejc (mentor), Stražar, Klemen (komentor), Fošnarič, Miha (recenzent)

Uvod: Pri predoperativnem načrtovanju zahtevnih ortopedskih posegov, kot so periacetabularna osteotomija (PAO), osteoplastika stegnenice ali acetabuloplastika, je natančna prostorska predstava kolčnega sklepa ključnega pomena. Poleg kostnih struktur ima pomembno biomehansko in stabilizacijsko vlogo tudi acetabularni labrum, ki ga magnetna resonanca (MR) prikaže bolje kot računalniška tomografija (CT). Izziv predstavlja segmentacija in registracija MR modela labruma v CT model medenice. Namen: Namen magistrske naloge je bil ugotoviti, ali je registracija MR modela labruma v CT model medenice natančnejša, če se za referenčno kostno strukturo uporabi glava stegnenice ali acetabulum. Dodatno smo ocenili izvedljivost ročne segmentacije labruma iz MR podatkov ter primernost programa 3D Slicer za kvantitativno analizo registracij. Metode dela: Raziskava je temeljila na retrospektivno zbranih CT in MR podatkih štirinajstih kolkov pri enajstih pacientih z razvojno displazijo kolka (DDH) ali s femoroacetabularno utesnitvijo (FAI). Labrum smo ročno segmentirali iz več MR sekvenc z uporabo programa Ekliptik EBS ter registrirali v CT model medenice glede na dve različni kostni referenci (acetabulum in stegnenica). Natančnost registracije smo ocenili z Diceovim koeficientom podobnosti, Hausdorffovo razdaljo in premikom centra. Preverjena je bila tudi ponovljivost izračunov metrik v programu 3D Slicer. Statistično analizo smo izvedli v programu IBM SPSS. Rezultati: Povprečni Diceov koeficient je znašal 0,236 ± 0,139, premik centra med modeli približno 3,6 mm, Hausdorffove razdalje pa so bile zmerno visoke. Med registracijo z referenco na glavo stegnenice in registracijo z referenco na acetabulum nismo ugotovili pomembnih razlik v izračunanih metrikah. Vrstni red vnosa modelov v 3D Slicer ni vplival na rezultate. Razprava in zaključek: Segmentacija labruma iz MR podatkov je izvedljiva, integracija v CT modele pa tehnično možna. Ugotovili smo, da izbira med acetabulumom in glavo stegnenice kot referenco ne povzroči statistično značilnih razlik, kar potrjuje, da so pri registraciji mehkih tkiv omejitve posledica predvsem toge metode. Zato je priporočljiva dodatna vizualna validacija in razvoj netogih metod. Modeli so uporabni pri predoperativnem načrtovanju korektivnih posegov, še bolj pa ob uporabi enotnih slikovnih protokolov in več opazovalcih.

• Jezik: Slovenski jezik
• Ključne besede: magistrska dela, radiološka tehnologija, kolčni labrum, segmentacija, registracija, magnetna resonanca, računalniška tomografija, Ekliptik EBS, 3D Slicer
• Vrsta gradiva: Magistrsko delo/naloga
• Tipologija: 2.09 - Magistrsko delo
• Organizacija: ZF - Zdravstvena fakulteta
• Kraj izida: Ljubljana
• Založnik: [E. Černelič]
• Leto izida: 2025
• Št. strani: 57 str., [1] str. pril.
• PID: 20.500.12556/RUL-177552
• UDK: 616-07
• COBISS.SI-ID: 262849539
• Datum objave v RUL: 24.12.2025

Sekundarni jezik

• Jezik: Angleški jezik
• Naslov: The impact of reference bone structure selection on the accuracy of 3D MRI hip labrum model registration into a 3D CT pelvis model : master thesis

Izvleček

Introduction: Accurate three-dimensional visualization of the hip joint is essential for preoperative planning of complex orthopedic procedures such as periacetabular osteotomy (PAO), femoral osteoplasty, and acetabuloplasty. In addition to bony structures, the acetabular labrum plays an important biomechanical and stabilizing role, and magnetic resonance imaging (MRI) provides better visualization of the labrum than computed tomography (CT). One of the main challenges is the segmentation and registration of MRI-based labrum models into CT-based pelvic models. Purpose: The aim of this master’s thesis was to determine whether registration of the labrum into CT models of the pelvis is more accurate when the femoral head or the acetabulum is used as the reference structure. In addition, the feasibility of manual segmentation of the labrum from MRI data and the suitability of 3D Slicer for quantitative analysis of registrations were assessed. Methods: The study was based on retrospectively collected CT and MRI data of fourteen hips from eleven patients with developmental dysplasia of the hip (DDH) or femoroacetabular impingement (FAI). The labrum was manually segmented from multiple MRI sequences using Ekliptik EBS and registered into CT pelvic models with two different references (acetabulum and femoral head). Registration accuracy was evaluated using the Dice similarity coefficient, Hausdorff distance, and center displacement. The reproducibility of metric calculations in 3D Slicer was also tested. Statistical analysis was performed in IBM SPSS. Results: The mean Dice coefficient was 0.236 ± 0.139, Hausdorff distances were moderately high, and the average center displacement was approximately 3.6 mm. No significant differences in accuracy metrics were found between acetabulum- and femoral head–based registrations. The order of model input in 3D Slicer had no impact on the results. Discussion and conclusion: Segmentation of the labrum from MRI data proved feasible, and integration into CT models was technically achievable. The choice between acetabulum and femoral head as reference did not result in statistically significant differences, confirming that the main limitations arise from rigid registration methods applied to soft tissues. Additional visual validation and the development of non-rigid approaches are therefore recommended. The models are applicable for preoperative planning of corrective procedures, especially when supported by standardized imaging protocols and multiple observers.

• Ključne besede: master's theses, radiologic technology, acetabular labrum, segmentation, registration, magnetic resonance imaging, computed tomography, Ekliptik EBS, 3D Slicer