Methods and systems for automated segmentation of dense cell populations
AssigneeGeneral Electric Company
InventorsJens Rittscher, Rahul Bhotika, Kishore Mosaliganti, Dirk Ryan Padfield, Raghu Machiraju
Life of this patent
Estimating the term…
Abstract
Systems and methods for segmenting images comprising cells, wherein the images comprise a plurality of pixels; one or more three dimensional (3D) clusters of cells are identified in the images; and the 3D clusters of cells are automatically segmented into individual cells using one or more models.
Claims24
A method for segmenting images of cells to measure morphology or distribution of the cells within a multi-layered, multi-dimensional cell population, comprising the steps of:
- providing one or more images comprising a plurality of pixels;
- identifying one or more three dimensional (3D) clusters of cells in the images, wherein each 3D cluster of cells comprises multiple layers of cells;
- finding a z-component of a cluster center for one or more of the 3D clusters of cells which comprises:
- creating a max index image, wherein each pixel corresponds to an image slice containing a maximum value at an x, y location;
- generating a histogram of one or more max indices around a border of one or more of the clusters derived from a mask generated in a step of segmenting the cluster;
- and
- locating a mode of the histogram and identifying one or more centers of one or more of the clusters based on the mode;
- and
- automatically segmenting one or more of the 3D clusters of cells into one or more individual cells using one or more models.
A system for segmenting images of cells to measure morphology or distribution of the cells within a multi-layered, multi-dimensional human cell population, comprising:
- a storage device for at least temporarily storing one or more z-stack images each comprising a plurality of pixels;
- and
- a processor configured to:
- identify one or more three dimensional (3D) clusters of cells in said images, each 3D cluster of cells being multi-layered;
- find a z-component of a cluster center for one or more of the 3D clusters of cells by:
- creating a max index image, wherein each pixel corresponds to an image slice containing a maximum value at an x,y location;
- generating a histogram of one or more max indices around a border of one or more of said clusters derived from a mask generated in a step of segmenting the cluster;
- and
- locating a mode of said histogram and identifying one or more centers of one or more of said clusters based on the mode;
- and
- segment said 3D clusters of cells into individual cells using one or more automated models.
A method for segmenting images of cells to measure morphology or distribution of the cells within a multi-layered, multi-dimensional cell population, comprising the steps of:
- providing one or more images comprising a plurality of pixels;
- identifying one or more three dimensional (3D) clusters of cells in the images, wherein each 3D cluster of cells comprises multiple layers of cells;
- and
- automatically segmenting one or more of the 3D clusters of cells into one or more individual cells using one or more models, wherein the step of segmenting one or more of the cluster of cells into one or more individual cells comprises:
- using one or more priors corresponding to one or more cell characteristics;
- initializing one or more probabilistic models, wherein at least one of the probabilistic models comprises a shape-based model comprising a distance map;
- applying a distance map saturation mechanism to said one or more images;
- segmenting one or more nuclei of said cells using a watershed segmentation
- subsampling the distance map;
- processing the subsampled distance map;
- and
- supersampling said processed distance map.
Dependencies are inferred from the claim text.