Systems and methods for segmentation and analysis of 3D images
AssigneePerkinElmer Cellular Technologies Germany GmbH
InventorsPeet Kask, Kaupo Palo, Hartwig Preckel
Life of this patent
Estimating the term…
Abstract
Described herein are computationally efficient systems and methods for processing and analyzing two-dimensional (2D) and three-dimensional (3D) images using texture filters that are based on the Hessian eigenvalues (e.g., eigenvalues of a square matrix of second-order partial derivatives) of each pixel or voxel. The original image may be a single image or a set of multiple images. In certain embodiments, the filtered images are used to calculate texture feature values for objects such as cells identified in the image. Once objects are identified, the filtered images can be used to classify the objects, for image segmentation, and/or to quantify the objects (e.g., via regression).
Claims27
A method of three-dimensional image analysis, the method comprising:
- receiving, by a processor of a computing device, a three-dimensional (3D) image comprising voxels;
- applying, by the processor, a set of second-derivative filters to the 3D image, thereby producing a corresponding set of second-derivative images;
- applying, by the processor, a set of rotationally invariant 3D texture filters to the set of second-derivative images using a set of predefined characteristic directions in Hessian eigenvalue space, thereby producing a corresponding set of filtered images;
- identifying, by the processor, using at least one of the set of filtered images or the 3D image, a plurality of objects in the 3D image;
- and
- performing, by the processor and using the set of filtered images, at least one of:
- (a) classifying a portion of the plurality of identified objects in the 3D image;
- (b) segmenting the 3D image to separate a portion of the plurality of identified objects from the 3D image;
- or
- (c) determining a quality of interest value for a portion of the plurality of identified objects in the 3D image.
A method of two-dimensional image analysis, the method comprising:
- receiving, by a processor of a computing device, a two-dimensional (2D) image comprising pixels;
- applying, by the processor, a set of second-derivative filters to the 2D image, thereby producing a corresponding set of second-derivative images;
- applying, by the processor, a set of rotationally invariant 2D texture filters to the set of second-derivative images using a set of predefined characteristic directions, thereby producing a corresponding set of filtered images;
- identifying, by the processor, using at least one of the set of filtered images or the 2D image, a plurality of objects in the 2D image;
- and
- performing, by the processor and using the set of filtered images, at least one of:
- (a) classifying a portion of the plurality of identified objects in the 2D image;
- (b) segmenting the 2D image to separate a portion of the plurality of identified objects from the 2D image;
- or
- (c) determining a quality of interest value for a portion of the plurality of identified objects in the 2D image.
A system for three-dimensional image analysis, the system comprising:
- a processor of a computing device;
- and
- a memory storing instructions that, when executed by the processor, cause the processor to:
- (i) receive a three-dimensional (3D) image comprising voxels;
- (ii) apply a set of second-derivative filters to the 3D image, thereby producing a corresponding set of second-derivative images;
- (iii) apply a set of rotationally invariant 3D texture filters to the set of second-derivative images using a set of predefined characteristic directions, thereby producing a corresponding set of filtered images;
- (iv) identify using at least one of the set of filtered images or the 3D image, a plurality of objects in the 3D image;
- and
- (v) perform, using the set of filtered images, at least one of:
- (a) classifying a portion of the plurality of identified objects in the 3D image;
- (b) segmenting the 3D image to separate a portion of the plurality of identified objects from the 3D image;
- and
- (c) determining a quality of interest value for a portion of the plurality of identified objects in the 3D image.
A system for two-dimensional image analysis, the system comprising:
- a processor of a computing device;
- and
- a memory storing instructions that, when executed by the processor, cause the processor to:
- (i) receive a two-dimensional (2D) image comprising pixels;
- (ii) apply a set of second-derivative filters to the 2D image, thereby producing a corresponding set of second-derivative images;
- (iii) apply a set of rotationally invariant 2D texture filters to the set of second-derivative images using a set of predefined characteristic directions, thereby producing a corresponding set of filtered images;
- (iv) identify using at least one of the set of filtered images or the 2D image, a plurality of objects in the 2D image;
- and
- (v) perform, using the set of filtered images, at least one of:
- (a) classifying a portion of the plurality of identified objects in the 2D image;
- (b) segmenting the 2D image to separate a portion of the plurality of identified objects from the 2D image;
- or
- (c) determining a quality of interest value for a portion of the plurality of identified objects in the 2D image.
Dependencies are inferred from the claim text.