Name
Title | ||
|---|---|---|
Improving the Robustness in Extracting 3D Point Landmarks from 3D Medical Images Using Parametric Deformable Models |
Abstract | ||
|---|---|---|
Existing approaches to the extraction of 3D point landmarks based on parametric deformable models suffer from their dependence on a good model initialization to avoid local suboptima during model fitting. Our main contribution to increasing the robustness of model fitting against local suboptima is a novel hybrid optimization algorithm combining the advantages of both the conjugate gradient (cg-)optimization method (known for its time efficiency) and genetic algorithms (exhibiting robustness against local suboptima). It has to be stressed, however, that the scope of applicability of this nonlinear optimization method is not restricted to model fitting problems in medical image analysis. We apply our model fitting algorithm to 3D medical images depicting tip-like and saddle-like anatomical structures such as the horns of the lateral ventricles in the human brain or the zygomatic bone as part of the skull. Experimental results for 3D MR and CT images demonstrate that in comparison to a purely local cg-optimization method, the robustness of model fitting in the case of poorly initialized model parameters is significantly improved with a hybrid optimization strategy. Moreover, we compare an edge strength-based fitting measure with an edge distance-based fitting measure w.r.t. their suitability for model fitting. |
Year | DOI | Venue |
|---|---|---|
2001 | 10.1007/3-540-45468-3_70 | MICCAI |
Keywords | Field | DocType |
fitting problem,fitting measure w,model fitting,initialized model parameter,model fitting algorithm,local suboptima,point landmarks,parametric deformable model,good model initialization,hybrid optimization strategy,fitting measure,parametric deformable models,conjugate gradient,nonlinear optimization,genetic algorithm | Conjugate gradient method,Fitting algorithm,Computer vision,Pattern recognition,Computer science,Nonlinear programming,Robustness (computer science),Parametric statistics,Artificial intelligence,Anatomical structures,Initialization,Genetic algorithm | Conference |
ISBN | Citations | PageRank |
3-540-42697-3 | 7 | 0.60 |
References | Authors | |
16 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Manfred Alker | 1 | 10 | 1.35 |
| Sönke Frantz | 2 | 76 | 8.01 |
| Karl Rohr | 3 | 377 | 52.96 |
| H. Siegfried Stiehl | 4 | 516 | 67.10 |