MSMI Segmentation

Using InShape models for the segmentat- ion of (medical) images

The basis software framework as well as the basis methods for the segmentationof medical images was developed in the course of the project "Modular model based segmentation framework"(MMS)". For more details on the developed methods and algorithms that have been developed at the IBIA and MSMI, please also refer to the corresponding publications.

Cardiac segmentation (Cine MRI):

3D shape model illustrating endocardial shape variations of the cardiac ventricles and atria

 

  

Segmentation result for endocardium using a 3D statistical model based on deformation fields in combination with level sets

 

 

3D myocardium segmentation (MRI)

Segmentation result for myocardium using statistical model based on  deformation fields in combination with level sets

 

 

 

4D endocardium segmentation

Statistical 4D InShape model of endocarium movement between 0ms and 100 ms of cardiac cycle (surface rendering, colors indicate the amount of deviation from the mean shape)

 

 

Statistical 4D InShape model of endocarium movement between 0 ms and 100 ms of cardiac cycle (volume rendering)

 

 

Endocardium segmentation result for 3 different datasets using 4D InShape model

  

Femur segmentation (CT)

3D InShape model of proximal femur (left: shape variations and average shape (colors indicate the amount of deviation from the mean shape); right: appearance variations + average appearance)

 

 

Segmenation result using 3D InShape model for segmentation of proximal femur

 

  

Segmenation result using 3D InShape model for segmentation of proximal femur in CT image containing "artifacts" (green contour illustrates initial surface,  white line illustrates segmentation result (2D example slice))

 

 

Vertebra segmentation (CT images)

InShape model of lumbar vertebra (shape component, colors indicate the amount of variation compared to the mean shape)

 

 

Segmentation result using 3D InShape model