Bayesian Programming

Probability as an extension of logic

Cognitive ScienceResearch

3D human hip volume reconstruction.

with incomplete multimodal medical images.

Application to computer assisted surgery of the total hip replacement (THR).

PhD thesis of Miriam Amavizca (2005)

This thesis is situated within the context of total hip prosthesis replacement (THR) computer-assisted surgery systems. The goal of this work is to construct a 3D hip model of the patient from partial data.

The standard THR planning process simply relies on X-ray image whereas THR computer-assisted systems use a 3D hip volume reconstructed from medical ima- gining techniques such as MRI, CT or Scanner. Nevertheless, 3D volumes reconstruction presents some disadvantages : (i) there is a long waiting time to get the imagining study , (ii) due to their cost, machines are not available in all clinics, (iii) a complete automatic segmentation process for the 3D volume constructions is not jet available and (iv) some patients can not be exposed to some imagining techniques such as MRI, CT or Scanner studies.

Considering these problems, we make two main contributions in this work :

  • –  a study of the exploitable characteristics of hip and femur for 3D model recons- truction,
  • –  a methodology for 3D hip volume reconstruction using minimally invasive ima- gining techniques : a single radiographic image (2D data) and a few echographic images (3D data).

The proposed method consist of three main stages : (i) data acquisition of the radiographic and echographic images of the patient hip, (ii) inference of the hip atlas of the patient and (iii) 3D hip volume reconstruction by a mesh deformation that adapts to the infered atlas. These stages pose different problems related to the representation of the generic atlas, to the inference process, and to the radiographic and echographic data processing. To solve this problematic we use Bayesian techniques.


Leave a Reply

Your email address will not be published. Required fields are marked *