Medical imaging apparatus, image processing apparatus, and image processing method

Patent metadata

Field	Value
Publication number	US-11450003-B2
Application number	US-201916594116-A
Country	US
Kind code	B2
Filing date	Oct 7, 2019
Priority date	Oct 29, 2018
Publication date	Sep 20, 2022
Grant date	Sep 20, 2022

Abstract

Provided is a technology for extracting an image of a target plane from 2D or 3D image data acquired by a medical imaging apparatus with a small amount of computation and at high speed. A plane of a target plane including a predetermined structure is extracted from image data of a subject. A region of the predetermined structure included in the plane is detected by applying a learning model learned using learning data including a target plane for learning including an image of the structure and a region-of-interest plane for learning obtained by cutting out and enlarging a partial region including the structure in the target plane for learning to a plurality of planes obtained from the image data, and the plane of the target plane is extracted based on the detected region of the predetermined structure.

First claim

What is claimed is:

1. A medical imaging apparatus comprising:
an imaging unit that collects image data of a subject; and
an image processing unit that performs a process of extracting, from the image data collected by the imaging unit, a plane of a target plane that includes a predetermined structure,
wherein the image processing unit includes a learning model storage unit that stores a learning model learned using learning data, a structure extraction unit that detects a region of the predetermined structure included in the plane by applying the learning model to a plurality of planes obtained from the image data, and a plane extraction unit that extracts the plane of the target plane from the plurality of planes based on the detected region of the predetermined structure,
wherein the learning data includes a target plane for learning including an image of the structure imaged in advance for the target plane, and a region-of-interest plane for learning obtained by cutting out and enlarging a region of interest including the structure in the target plane for learning,
wherein the target plane includes, as the predetermined structure, a first structure and a second structure located within the first structure,
wherein the target plane for learning includes images of the first structure and the second structure,
wherein the region-of-interest plane for learning is an image obtained by cutting out and enlarging a region of the target plane that includes the first structure of the target plane for learning,
wherein when a region of the first structure is detected by applying the learning model to the plane obtained from the image data, the structure extraction unit generates an image obtained by cutting out a region including the region of the first structure from the plane, applies the learning model to the cut-out image, and detects a region of the second structure,
wherein the image processing unit further includes a total score computation unit,
wherein the structure extraction unit further outputs a first score indicating a reliability of detection for the detected region of the first structure of the plane and a second score indicating a reliability of detection for the detected region of the second structure of the plane,
wherein the total score computation unit computes:
a geometric score from a geometric positional relationship between regions of the structures detected by the structure extraction unit, the geometric positional relationship being an angle between two regions of structures, the geometric score being calculated based on the angle and a weight value, and
a total score which is a summation of the first score indicating the reliability of detection for the detected region of the first structure of the plane, the second score indicating the reliability of detection for the detected region of the second structure of the plane, and the geometric score,
wherein the plane extraction unit selects the plane, the total score of which is relatively high, as a plane of the target plane,
wherein the structure extraction unit successively extracts regions of the structure for a plurality of planes obtained from the image data,
wherein the total score computation unit computes the total score for each of the planes, and
wherein the plane extraction unit selects a tomographic image of the target plane by analyzing changes in the total score successively computed for the plurality of planes, and causes a display unit to display the selected tomographic image.

2. The medical imaging apparatus according to claim 1 , wherein when a region of a structure included in the plane is detected by applying the learning model to the plane obtained from the image data, the structure extraction unit generates an image obtained by cutting out a region including the region of the structure from the plane and further detects a structure included in the cut-out image by applying the learning model again.

3. The medical imaging apparatus according to claim 1 , wherein the learning data further includes information specifying a type of the structure and position information of a region in which the structure is located on an image, each of which is associated with the target plane for learning or the region-of-interest plane for learning including the structure.

4. The medical imaging apparatus according to claim 1 , wherein the learning model is a reduced model, in which at least one of an input image size and the number of levels of a high-accuracy model learned using the target plane as learning data is reduced based on analysis of a relative size of a detection target region, relearned using the learning data including the target plane for learning and the region-of-interest plane for learning.

5. The medical imaging apparatus according to claim 4 , wherein when the learning model is the reduced model in which the input image size is reduced, the structure extraction unit reduces the plane obtained from the image data to the input image size and inputs the plane to the learning model as an input image.

6. The medical imaging apparatus according to claim 1 , wherein the image processing unit further includes an automatic measurement unit that computes a measurement value determined in advance using a region of the structure of the plane selected as a plane of the target plane.

7. The medical imaging apparatus according to claim 1 , wherein the image processing unit further includes a display control unit that causes a display unit to display a plane of the target plane selected by the plane extraction unit, a region of the structure, and the total score.

8. The medical imaging apparatus according to claim 1 , wherein the structure extraction unit selects one of a plurality of detected regions of the structure as a region of interest.

9. The medical imaging apparatus according to claim 8 , further comprising
a display control unit that causes a display unit to enlarge and display a region of the plane corresponding to the region of interest selected by the structure extraction unit.

10. The medical imaging apparatus according to claim 8 ,
wherein the imaging unit collects the image data by transmitting an ultrasonic beam while scanning a subject, and
the imaging unit limits a scanning range of the ultrasonic beam to a range of the region of interest selected by the structure extraction unit.

11. An image processing apparatus for performing a process of receiving image data from a subject and extracting, from the image data, a plane of a target plane that includes a predetermined structure, the image processing apparatus comprising:
a learning model storage unit that stores a learning model learned using learning data;
a structure extraction unit that detects a region of the predetermined structure included in the plane by applying the learning model to a plurality of planes obtained from the image data; and
a plane extraction unit that extracts the plane of the target plane based on the detected region of the predetermined structure,
wherein the learning data includes a target plane for learning including an image of the structure captured in advance for the target plane, and a region-of-interest plane for learning obtained by cutting out and enlarging a partial region including the structure in the target plane for learning,
wherein the target plane includes, as the predetermined structure, a first structure and a second structure located within the first structure,
wherein the target plane for learning includes images of the first structure and the second structure,
wherein the region-of-interest plane for learning is an image obtained by cutting out and enlarging a region of the target

Assignees

• Hitachi Ltd
• Fujifilm Healthcare Corp

Inventors

• Li Yun
• Toyomura Takashi
• Inoue Kenta
• Maeda Toshinori

Classifications

• G06T12/00

  Tomographic reconstruction from projections · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06T2207/10081

  Computed x-ray tomography [CT] · CPC title