System and method for retrieval of similar findings from a hybrid image dataset

Patent metadata

Field	Value
Publication number	US-10176612-B2
Application number	US-201514751708-A
Country	US
Kind code	B2
Filing date	Jun 26, 2015
Priority date	Jun 27, 2014
Publication date	Jan 8, 2019
Grant date	Jan 8, 2019

Abstract

In a method for retrieval of similar findings from a hybrid image dataset, a database of hotspots is prepared, wherein the hotspots are identified by binary strings encoding descriptors, and identify binary strings stored in the database are identified that resemble a new binary string.

First claim

We claim as our invention:

1. A method for retrieval of similar findings from a hybrid image dataset, comprising:
in a processor, preparing a database of local maxima, identified by binary strings encoding descriptors;
providing said processor with functional image data representing a functional image of a subject, and anatomical image data representing am anatomical image of the subject, said functional image containing at least one local maximum having an intensity that is above a clinically-relevant threshold value;
in said processor, aligning said functional image data and said anatomical image data to form the hybrid image dataset;
in said processor, normalizing the functional image data within the hybrid image dataset;
in said processor, identifying each local maximum in said functional image data;
in said processor, segmenting regions of the functional image, with each region of interest containing a respective local maximum in order to designate regions of interest in said functional image;
in said processor, defining a bounding box, having a size and shape, around each region of interest in said functional image and then defining an equivalent bounding box, having the same size and shape, for an image region in the anatomical image data aligned with the functional image data in the hybrid image dataset;
in said processor, processing each image region of the hybrid image dataset, corresponding to the location of the equivalent bounding box, to calculate descriptors for the image region bounded by the equivalent bounding box in the hybrid image dataset;
in said processor, encoding descriptors relating to the processed image region into a new binary string;
in said processor, comparing the new binary string to a number of binary, strings stored in the database; and
in said processor, identifying binary strings stored in the database which resemble the new binary string, and making the new binary string available in electronic format output of the processor for retrieval of said similar findings.

2. A method according to claim 1 , wherein the encoded descriptors comprise image intensity descriptors.

3. A method according to claim 2 , comprising identifying local maxima by identifying regions with an intensity above a certain clinically-relevant threshold in the functional image.

4. A method according to claim 1 , wherein the encoded descriptors comprise localization descriptors.

5. A method according to claim 4 comprising defining localization descriptors relative to a set of anatomical landmarks.

6. A method according to claim 4 wherein defining the localization descriptors as a set of vector displacements from the anatomical landmarks.

7. A method according to claim 1 comprising displaying image regions corresponding to the identified binary strings to a user.

8. A method according to claim 1 comprising using the binary strings stored within the database to encode descriptors relating to pre-processed image regions defined by a bounding box of the same dimension as the bounding box defining the image region used to define the new binary string.

9. A method according to claim 1 comprising evaluating which binary strings are similar based on the encoded descriptors.

10. A method according to claim 1 comprising storing image regions corresponding to the identified binary strings, and processing the stored image regions for presentation to a user.

11. A method according to claim 10 comprising, in response to user selection of a local maximum, retrieving image regions corresponding to the identified binary strings and presented to the user for comparison, along with image region corresponding to the selected local maximum.

12. A method according to claim 1 , wherein the database further contains annotations linked with image regions, and displaying said annotations to a user together with the image region.

13. An apparatus for retrieval of similar findings from a hybrid image dataset, comprising:
a processor in communication with a display device;
said processor being provided with functional image data representing a functional image of a subject, and anatomical image data representing am anatomical image of the subject, said functional image containing at least one local maximum having an intensity that is above a clinically-relevant threshold value;
said processor being configured to align said functional image data and said anatomical image data to form the hybrid image dataset;
said processor being configured to normalize the functional image data within the hybrid image dataset;
said processor being configured to identify each local maximum in said functional image data;
said processor being configured to segment regions of the functional image in order to designate regions of interest in said functional image, with each region of interest containing a respective local maximum;
said processor being configured to define a bounding box, having a size and shape, around each region of interest in said functional image and then to define an equivalent bounding box, having the same size and shape, for an image region in the anatomical image data aligned with the functional image data in the hybrid image dataset;
said processor being configured to process the image region of the hybrid image dataset corresponding to the location of the equivalent bounding box, to calculate image-intensity descriptors and localization descriptors for the image region bounded by the equivalent bounding box in the hybrid image dataset;
said processor being configured to encode the localization descriptors and image intensity descriptors for the image region into a new binary string;
said processor being configured to compare the new binary string to a number of binary strings stored in the database and identify binary strings stored in a database of local maxima which resemble the new binary string; and
said processor being configured to cause the multi-fused image of the regions to be displayed at said display device.

14. A non-transitory, computer-readable data storage medium encoded with programming instructions, said storage medium being loaded into a processor that is in communication with a display device, and said programming instructions causing said processor to:
receive functional image data representing a functional image of a subject, and anatomical image data representing am anatomical image of the subject;
align said functional image data and said anatomical image data to form the hybrid image dataset, said functional image containing at least one local maximum having an intensity that is above a clinically-relevant threshold value;
normalize the functional image data within the hybrid image dataset;
identify each local maximum in said functional image data;
segment regions of the functional image data in order to designate regions of interest in said functional image data, with each region of interest containing a respective local maximum;
define a bounding box, having a size and shape, around each region of interest in said functional image data then define an equivalent bounding box, having the same size and shape, for an image region in the anatomical image data aligned with the functional image data in the hybrid image dataset;
process the image region of the hybrid image dataset corresponding to the location of the equivalent bounding box, to calculate image-intensity descriptors and localization descriptors for the image region bounded by the equivalent bounding box in the hybrid image dataset;
encode the localization descriptors and image intensity descriptors for the image region into a new binary

Assignees

• Siemens Medical Solutions Usa Inc

Inventors

• Kelly Matthew David
• Schottlander David
• Sibille Ludovic

Classifications

• G06V10/464

  using a plurality of salient features, e.g. bag-of-words [BoW] representations · CPC title

• G06V10/758

  Involving statistics of pixels or of feature values, e.g. histogram matching · CPC title

• G06T2207/10108

  Single photon emission computed tomography [SPECT] · CPC title

• G06T2207/10104

  Positron emission tomography [PET] · CPC title

• G06T2207/30096

  Tumor; Lesion · CPC title