Search

DE-102020209416-B4 - Method for deformation simulation and device

DE102020209416B4DE 102020209416 B4DE102020209416 B4DE 102020209416B4DE-102020209416-B4

Abstract

Method for deformation simulation of a hollow organ deformable by the insertion of a medical instrument or medical object (10) comprising the following steps: • Provision of a pre-trained machine learning algorithm (24), • Provision of a medical 3D image of the hollow organ (10) with surrounding tissue, which 3D image was taken prior to the insertion of a medical instrument (3; 4) or medical object, • Segmentation or provision of a segmentation of the medical 3D image of the hollow organ (10) and determination or provision of a three-dimensional model of the hollow organ (10), • Providing information about an imported or to-be-introduced medical instrument (3; 4) or medical object, and • Simulation of the expected or caused deformation of the hollow organ (10) by the introduction of the instrument (3; 4) or medical object based on the segmented medical 3D image of the hollow organ (10) and the surrounding tissue and information about the instrument or object by using the pre-trained machine learning algorithm (24).

Inventors

  • Katharina Breininger
  • Marcus Pfister

Assignees

  • Siemens Healthineers Ag

Dates

Publication Date
20260513
Application Date
20200727
Priority Date
20190731

Claims (20)

  1. Method for deformation simulation of a hollow organ (10) deformable by the insertion of a medical instrument or medical object, comprising the following steps: • Provision of a pre-trained machine learning algorithm (24), • Provision of a medical 3D image of the hollow organ (10) with surrounding tissue, which 3D image was acquired prior to the insertion of a medical instrument (3; 4) or medical object, • Segmentation or provision of a segmentation of the medical 3D image of the hollow organ (10) and determination or provision of a three-dimensional model of the hollow organ (10), • Provision of information about an inserted or to-be-inserted medical instrument (3; 4) or medical object, and • Simulation of the expected or caused deformation of the hollow organ (10) by the insertion of the instrument (3; 4) or medical object based on the segmented medical 3D image of the hollow organ (10) and the surrounding tissue and the information about the instrument or object by using the pre-trained machine learning algorithm (24).
  2. Procedure according to Claim 1 , where the model is formed from a surface or volume model.
  3. Procedure according to Claim 1 or 2 , wherein the machine learning algorithm (24) for the simulation of the deformation of the hollow organ takes into account information about the influence of tissue properties of the surrounding tissue of the hollow organ (10).
  4. Method according to one of the preceding claims, wherein the machine learning algorithm mus (24) is pre-trained by means of a multitude of known image pairs (MO; MM) from a first medical 3D image of an undistorted hollow organ and its surrounding tissue and a second medical 3D image of a hollow organ (13) and its surrounding tissue deformed by a medical instrument (3; 4) or medical object.
  5. Procedure according to Claim 4 , wherein the recording pairs (MO; MM) are each segmented with respect to the hollow organ (10) and the hollow organ (10) is represented as a three-dimensional model with sub-elements, in particular as a grid with grid elements.
  6. Procedure according to Claim 5 , wherein the machine learning algorithm (24) for the deformation simulation of the hollow organ (10) takes into account stiffness parameters (S ij ) between sub-elements, which stiffness parameters (S ij ) are taken from the medical 3D image as parameterized functions of the tissue properties of the surrounding tissue of the respective sub-elements.
  7. Procedure according to Claim 6 , where those stiffness parameters (S ij ) are determined which minimize the error between real deformation and deformation simulation.
  8. Method according to one of the preceding claims, wherein the properties of the tissue surrounding the model are mapped onto the model, wherein the properties in particular influence the local deformation behavior of the hollow organ (10).
  9. Procedure according to one of the Claims 5 until 8 , wherein the surrounding tissue of a sub-element is mapped onto the sub-element by a vector (V) of predetermined length.
  10. Procedure according to one of the Claims 5 until 8 , wherein the surrounding tissue of a sub-element is mapped onto the sub-element by a hemispherical perimeter with a predetermined radius.
  11. Method according to any of the preceding claims, wherein it is assumed that tissue properties are represented by Hounsfield units (HU).
  12. Procedure according to one of the Claims 5 until 11 , wherein the sub-elements are designed as polygons, splines or other suitable mathematical formulations.
  13. Procedure according to Claim 4 or 5 , wherein the algorithm (24) is trained on the recording pairs (MO; MM) by minimizing the error function as the error between actual and simulated deformation.
  14. Procedure according to Claim 4 or 5 , where the algorithm is trained on the pairs of images (MO; MM) by replacing the second medical 3D image with a second, physically motivated simulation, in particular by FE methods.
  15. Method according to one of the preceding claims, wherein the deformation simulation is displayed on a display device (22).
  16. Procedure according to Claim 15 , wherein the deformation simulation is superimposed with at least one 2D recording, in particular a live X-ray recording.
  17. Method according to one of the preceding claims, wherein the information about the medical instrument (3; 4) to be introduced or already introduced is taken from a 2D image, in particular a radiographic image.
  18. Method according to one of the preceding claims, wherein the information about the medical instrument (3; 4) or medical object to be introduced takes into account its planned position.
  19. Method according to one of the preceding claims, wherein a deformation of the medical instrument or object is additionally simulated.
  20. System for performing a deformation simulation of a hollow organ (10) of a patient that can be deformed by the insertion of a medical instrument (3; 4) or medical object after one of the Claims 1 until 19 , comprising • a communication device (18) for querying medical 3D images of the hollow organ with surrounding tissue, • a storage device (19) for storing medical 3D images of the hollow organ with surrounding tissue, • an image processing device (20) for performing segmentation of medical 3D images of the hollow organ (10) with surrounding tissue and for determining a three-dimensional model of the surface of the hollow organ (10) with surrounding tissue, • a pre-trained machine learning algorithm (24) which is trained to perform the by introducing to simulate the expected or caused deformation of the hollow organ (10) by the instrument on the basis of a segmented medical 3D image, • a computing device (21) with a processor for executing the pre-trained machine learning algorithm (24), and • a display device (22) for displaying the modeled hollow organ (13) deformed by the instrument or object.

Description

The invention relates to a method for deformation simulation of a hollow organ deformable by the insertion of a medical instrument according to claim 1 and a device for carrying out such a method according to claim 20. This is particularly relevant in the field of interventional treatment (e.g., repair) of aortic aneurysms. Specifically, it concerns the nonlinear adaptation of the vessels to inserted rigid instruments (such as guide wires, catheters, and stents), a so-called deformation correction. An abdominal aortic aneurysm 2 - see 1 An iliac aneurysm is a bulge in the abdominal aorta (1), the extension of which into the leg arteries is called an iliac aneurysm. It is treated either with open abdominal surgery or minimally invasively by inserting a so-called stent graft (3). This procedure is called EVAR = endovascular aneurysm repair. Guide wires (4) and catheters are inserted into the abdominal aorta (1) via the two groins, through which one or more stent grafts (3) (a combination of a stent and an artificial blood vessel) are introduced. In established state-of-the-art procedures, EVAR is performed on angiography systems under fluoroscopic guidance. To minimize the application of iodine-containing (nephrotoxic) contrast agents, various methods have been described for overlaying the fluoroscopically recorded preoperative datasets (mostly CT angiographies) onto the image. The CTA datasets are segmented beforehand for this purpose. Typically, significant deformation occurs in the highly curved iliac vessels (and also in the aortas) due to the insertion of rigid instruments (such as guidewires and catheters). These deformations depend primarily on the local characteristics of the vessel and the surrounding tissue and are therefore not homogeneous. For example, the vessel will deform less in calcified areas than in other areas. 4 A hollow organ is shown, in which strongly calcified regions 11 are indicated by closely spaced dashed lines. There are methods to compensate for this deformation in the superimposed tissue intraoperatively. For example, from the article by Toth et al., Adaptation of 3D Models to 2D X-ray Images during Endovascular Abdominal Aneurysm Repair, Proc. Of the MICCAI Workshop, 2015, pp. 339-346 , a method for determining the deformation of a vessel using superimposed image datasets is known. In the 2 and 3 Such an intraoperative deformation correction has been demonstrated, whereby the 2 The original hollow organ 10 and an inserted guide wire 4 are shown in superimposition. 3 Figure 13 shows the corrected deformed hollow organ. Furthermore, there are methods that simulate the deformation pre-operatively, e.g., from the article by Roy et al., Finite element analysis of abdominal aortic aneurysms: geometrical and structural reconstruction with application of an anisotropic material model, IMA J Appl Math, 79 (5), 2014, pp. 1011-1026 , where predetermined or measured properties of instruments and tissues are used for simulation. However, this is very complex and must be determined experimentally. From the US 2017/ 0189118 A1 is a method for modeling non-rigid tissue deformation during virtual navigation of an interventional instrument within the tissue. From the DE 10 2010 041735 A1 is a method for visualizing a vessel of a biological object with an interventional instrument inserted into the vessel. It is an object of the present invention to provide a method which ensures a particularly simple and robust simulation of the deformation caused by the insertion of a medical instrument into a hollow organ; furthermore, it is an object of the invention to provide a device suitable for carrying out the method. The problem is solved according to the invention by a method for deformation simulation of a hollow organ deformable by the insertion of a medical instrument or object according to claim 1 and by a device according to claim 20. Advantageous embodiments of the invention are the subject of the respective dependent claims. In the inventive method for deformation simulation of a hollow organ deformable by the insertion of a medical instrument or medical object, the following steps are performed: provision of a pre-trained machine learning algorithm, provision of a medical 3D image of the hollow organ with surrounding tissue, which 3D image was taken before the insertion of a medical instrument or medical object, segmentation or provision of a segmentation of the medical 3D image of the hollow organ and determination or provision of a three-dimensional model of the hollow organ, provision of information about an inserted or to-be-inserted medical medical instrument or object, and simulation of the deformation of the hollow organ caused or expected by the introduction of the instrument or object based on the segmented medical 3D recording of the hollow organ and surrounding tissue and information about the instrument or object by using the pre-trained machine learning algorithm. The inventive m