Apparatus and method for multi-atlas based segmentation of medical image data

Patent metadata

Field	Value
Publication number	US-9818200-B2
Application number	US-201314079908-A
Country	US
Kind code	B2
Filing date	Nov 14, 2013
Priority date	Nov 14, 2013
Publication date	Nov 14, 2017
Grant date	Nov 14, 2017

Abstract

An image data processing apparatus including a data receiver receiving image data to be segmented, and an atlas selection processor accessing a plurality of atlas data sets and selecting a subset of the atlas data sets for use in segmenting the image data, wherein the atlas selection processor is configured to select the subset of atlas data sets in dependence on the positions of one or more anatomical landmarks comprised in the plurality of atlas data sets.

First claim

The invention claimed is:

1. A medical image data processing apparatus comprising:
a data receiver receiving image data to be segmented; and
an atlas selection processor accessing a plurality of atlas data sets and selecting a subset of the atlas data sets for use in segmenting the image data; wherein
each atlas data set comprises image data comprising a set of pixels or voxels, a plurality of predetermined anatomical landmarks, and associated data that identifies positions of the anatomical landmarks; and
the atlas selection processor configured to:
for each atlas data set, obtain a registration between the set of pixels or voxels of the atlas data set and the image data to be segmented, using a first registration method in real time;
for each atlas data set, transform the positions of the anatomical landmarks represented by the associated data, in accordance with the registration for that atlas data set of the image data to be segmented;
for each atlas data set, compute a distance between the transformed positions of the anatomical landmarks of the atlas data set in a coordinate space of the image data to he segmented and the transformed positions of anatomical landmarks for other of the atlas data sets in the coordinate space of the image data to be segmented;
perform an on-line clustering method, following the first registration method, comprising forming a plurality of clusters of the atlas data sets based on the values of the computed distances between the transformed positions of the anatomical landmarks, and select one of the clusters based on the values of the computed distances between the transformed positions of the anatomical andmarks for the clusters; and
select the atlas data sets of the selected cluster to be said subset of atlas data sets for use in segmenting the image data.

2. The image data processing method according to claim 1 , wherein the atlas selection processor is configured to perform a. second registration between the subset of atlas data sets and the image data to he segmented, using a second registration method that differs from the first registration method.

3. The image data processing apparatus according to claim 2 , wherein the first registration comprises a less computationally intensive registration method than that used for the second registration.

4. The image data processing apparatus according to claim 2 , wherein the second registration comprises a non-rigid registration.

5. The image data processing apparatus according to claim 1 , wherein the first registration comprises a rigid registration.

6. The image data processing apparatus according to claim 1 , wherein the atlas selection processor is configured to calculate, for each pair of atlas data sets, a respective distance metric between the transformed landmarks of one of the pair of the atlas data sets and the transformed landmarks of the other of the pair of the atlas data sets.

7. The image data processing apparatus according to claim 1 , wherein the atlas selection processor is configured to select the subset of atlas data sets by identifying atlas data sets that have the most similar distance metrics between the transformed landmarks of one of the pair of the atlas data sets and the transformed landmarks of the other of the pair of the atlas data sets.

8. The image data processing apparatus according to claim 1 , wherein computing the distance metric comprises determining a mean of the distance between the transformed landmarks of one of the pair of the atlas data sets and the transformed landmarks of the other of the pair of the atlas data sets.

9. The image data processing apparatus according to claim 1 , wherein for each of the atlas data sets the distance metric is computed in the coordinate system of the image data to be segmented.

10. The image data processing apparatus according to claim 1 , wherein the position of the landmarks is the position of the landmarks in the coordinate system of the image data to be segmented.

11. The image data processing apparatus according to claim 1 , wherein the clustering method comprises hierarchical clustering or spectral clustering.

12. The image data processing apparatus according to claim 1 , wherein the clustering method comprises identifying a largest cluster for a given distance or identifying a first cluster to contain a given or preset number of the atlas data sets.

13. The image data processing apparatus according to claim 1 , wherein the clustering method comprises an iterative clustering method that comprises grouping two or more atlas data sets having the most similar distance metric at each iteration.

14. The image data processing apparatus according to claim 1 , wherein the atlas selection processor segments the image data using the selected subset of atlas data sets to produce segmented image data.

15. A method of image data processing comprising:
receiving image data;
accessing a plurality of atlas datasets for use in segmenting the image data; and
selecting a subset of the atlas data sets for use in segmenting the image data, wherein
each atlas data set comprises image data comprising a set of pixels or voxels, a plurality of predetermined anatomical landmarks, and associated data that identifies positions of the anatomical landmarks; and
the selecting of the subset of the atlas data sets comprises;
for each atlas data set, obtaining a registration between the set of pixels or voxels of the atlas data set and the image data to be segmented, using a first registration method in real time:
for each atlas data set, transforming the positions of the anatomical landmarks represented by the associated data, in accordance with the registration for that atlas data set of the image data to be segmented;
for each atlas data set, computing a distance between the transformed positions of the anatomical landmarks of the atlas data set in a coordinate space of the image data to be segmented and the transformed positions of anatomical landmarks for other of the atlas data sets in the coordinate space of the image data to be segmented:
performing an on-line clustering method, following the first registration method comprising forming a plurality of clusters of the atlas data sets based on the values of the computed distances between the transformed positions of the anatomical landmarks, and selecting one of the clusters based on the values of the computed distances between the transformed positions of the anatomical landmarks for the clusters; and
selecting the atlas data sets of the selected cluster to be said subset of atlas data sets for use in segmenting the image data.

16. A non-transitory computer readable medium comprising a computer program product storing computer-readable instructions that are executable to perform a method according to claim 15 .

Assignees

• Toshiba Medical Sys Corp

Inventors

• Mohr Brian
• Poole Ian

Classifications

• G06T2207/10136

  3D ultrasound image · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title

• G06T7/174 Primary

  involving the use of two or more images · CPC title

• G06T2207/30016

  Brain · CPC title

• G06T2207/20128

  Atlas-based segmentation · CPC title