System and method for generating three dimensional geometric models of anatomical regions

Patent metadata

Field	Value
Publication number	US-12045929-B2
Application number	US-202117793077-A
Country	US
Kind code	B2
Filing date	Jan 14, 2021
Priority date	Jan 24, 2020
Publication date	Jul 23, 2024
Grant date	Jul 23, 2024

Abstract

A three-dimensional geometric image of an anatomical region is generated from a plurality of two-dimensional echographic image slices of the region. The image slices are filtered using a reaction-diffusion partial differential equation model before being arranged into a voxel space. Each voxel is then assigned a voxel value to create a volumetric data set from which the volumetric image can be rendered. The image is rendered from far to near, relative to a preset viewing direction, by an alpha-blending process. The alpha value at any given voxel can be determined using the magnitude of the density gradient vector at that voxel. Similarly, the direction of the density gradient vector at a given voxel can be used as a surface normal vector for shading purposes at that voxel.

First claim

What is claimed is:

1. A method of generating a three-dimensional image of an anatomical region from a plurality of two-dimensional echographic image slices of the anatomical region, wherein each image slice of the plurality of image slices is associated with localization information, the method comprising:
defining a three-dimensional voxel space comprising a plurality of voxels;
filtering the plurality of image slices using a reaction-diffusion partial differential equation model;
arranging the filtered plurality of image slices into the voxel space using the associated localization information;
assigning each voxel of the plurality of voxels a voxel value using the arranged, filtered plurality of image slices, thereby creating a volumetric data set; and
rendering a three-dimensional volumetric image from the three-dimensional data set, wherein the rendering comprises:
dividing the voxel space into a plurality of slices perpendicular to a preset viewing direction; and
alpha-blending the plurality of slices according to a process comprising, for each voxel within the plurality of voxels:
computing a density gradient vector;
computing a magnitude of the density gradient vector; and
assigning an alpha value using the magnitude of the density gradient value according to a process comprising:
normalizing the magnitude of the density gradient vector; and
exponentiating the normalized magnitude of the density gradient vector by an opacity bias.

2. The method according to claim 1 , wherein assigning each voxel of the plurality of voxels a voxel value using the arranged, filtered plurality of image slices comprises:
assigning each voxel of the plurality of voxels a greyscale value;
defining a neighborhood size; and, for each voxel within the plurality of voxels:
assigning the voxel a greyscale array, wherein the greyscale array is defined by the respective greyscale values of a neighborhood of the plurality of voxels within the neighborhood size of the voxel;
quantizing the greyscale array to a preset number of buckets; and
assigning the voxel value to the voxel according to the quantized greyscale array.

3. The method according to claim 2 , wherein assigning the voxel value to the voxel according to the quantized voxel array comprise assigning the voxel value to the voxel according to a majority vote of the quantized voxel array.

4. The method according to claim 1 , wherein filtering the plurality of image slices using a reaction-diffusion partial differential equation model comprises filtering the plurality of image slices using an Allen-Cahn-type reaction-diffusion model.

5. The method according to claim 1 , wherein dividing the voxel space into a plurality of slices perpendicular to a preset viewing direction occurs from a most distant slice of the plurality of slices to a nearest slice of the plurality of slices, as determined relative to the preset viewing direction.

6. The method according to claim 1 , wherein the opacity bias comprises a user preset value.

7. The method according to claim 1 , wherein the opacity bias comprises a scaled opacity bias.

8. The method according to claim 1 , wherein rendering the three-dimensional volumetric image from the three-dimensional data set further comprises, for each voxel within the plurality of voxels:
computing a density gradient vector;
computing a direction of the density gradient vector; and
defining the direction of the density gradient vector as a surface normal vector for shading purposes.

9. A method of generating a three-dimensional geometric image of an anatomical region, the method comprising:
defining a three-dimensional voxel space comprising a plurality of voxels;
receiving a plurality of two-dimensional echographic image slices of the anatomical region, wherein each image slice of the plurality of image slices is associated with localization information;
arranging the plurality of image slices into the voxel space using the associated localization information, thereby creating a volumetric data set; and
rendering a three-dimensional volumetric image from the three-dimensional data set,
wherein rendering the three-dimensional volumetric image from the three-dimensional data set comprises, for each voxel within the plurality of voxels:
computing a density gradient vector;
computing a magnitude of the density gradient vector; and
assigning an alpha value for alpha-blending purposes using the magnitude of the density gradient vector, wherein the assigning comprises:
normalizing the magnitude of the density gradient vector; and
exponentiating the normalized magnitude of the density gradient vector by an opacity bias.

10. The method according to claim 9 , wherein the opacity bias comprises a user preset value.

11. The method according to claim 9 , wherein the opacity bias comprises a scaled opacity bias.

12. The method according to claim 9 , wherein rendering the three-dimensional volumetric image from the three-dimensional data set further comprises:
computing a direction of the density gradient vector; and
defining the direction of the density gradient vector as a surface normal vector for shading purposes.

13. The method according to claim 9 , further comprising, prior to arranging the plurality of image slices into the voxel space using the associated localization information, thereby creating the volumetric data set, filtering the plurality of image slices using a reaction-diffusion partial differential equation model.

14. The method according to claim 13 , wherein the reaction-diffusion partial differential equation model comprises an Allen-Cahn-type reaction-diffusion model.

15. The method according to claim 9 , wherein arranging the plurality of image slices into the voxel space using the associated localization information, thereby creating the volumetric data set, comprises, for each voxel of the plurality of voxels:
assigning the voxel a greyscale value;
defining a neighborhood size;
assigning the voxel a greyscale array, wherein the greyscale array is defined by the respective greyscale values of a neighborhood of the plurality of voxels within the neighborhood size of the voxel;
quantizing the greyscale array to a preset number of buckets; and
assigning a final voxel value to the voxel according to the quantized greyscale array.

16. A system for generating a three-dimensional geometric image of an anatomical region, comprising:
an imaging and modeling module configured to:
define a three-dimensional voxel space comprising a plurality of voxels;
receive a plurality of two-dimensional echographic image slices of the anatomical region, wherein each image slice of the plurality of image slices is associated with localization information;
arrange the plurality of image slices into the voxel space using the associated localization information, thereby creating a volumetric data set;
compute a density gradient vector for each voxel within the plurality of voxels; and
render a three-dimensional volumetric image from the three-dimensional data set using a magnitude of the density gradient vector to assign an alpha value for alpha-blending purposes and a direction of the density gradient vector as a surface normal vector for shading purposes, wherein using the magnitude of the density gradient value to assign the alpha value comprises:
normalizing the magnitude of the density gradient vector; and
exponentiating the normalized magnitude of the density gradient vector by an opacity bias.

Assignees

• St Jude Medical Cardiology Div Inc

Inventors

• Dascal Lorina
• Sudarsky Oded
• Segev Jacob
• Dathskovsky Alexander
• Valach Yaron
• Yoresh Maxim

Classifications

• G06T2210/41

  Medical · CPC title

• G06T19/20

  Editing of three-dimensional [3D] images, e.g. changing shapes or colours, aligning objects or positioning parts · CPC title

• A61B8/5246

  combining images from the same or different imaging techniques, e.g. color Doppler and B-mode · CPC title

• A61B8/483

  involving the acquisition of a 3D volume of data · CPC title

• A61B8/5223

  for extracting a diagnostic or physiological parameter from medical diagnostic data (for algorithms to analyse biomedical images G06T7/0012) · CPC title