Method and system for automatic calcium scoring from medical images

Patent metadata

Field	Value
Publication number	US-12307660-B2
Application number	US-202117505917-A
Country	US
Kind code	B2
Filing date	Oct 20, 2021
Priority date	Mar 5, 2021
Publication date	May 20, 2025
Grant date	May 20, 2025

Abstract

Embodiments of the disclosure provide systems and methods for medical image analysis. A method may include receiving a medical image acquired of a subject by an image acquisition device. The method may also include applying a calcium detection model to detect at least one calcium region relevant in determining a calcium score from the medical image. The method may further include applying a score regression learning model to the at least one calcium region to determine a calcium score for the medical image. The method may additionally include providing the determined calcium score of the medical image for a diagnosis of the subject.

First claim

What is claimed is:

1. A system for determining a calcium score from a medical image, comprising:
a communication interface configured to receive the medical image acquired of a subject by an image acquisition device; and
at least one processor, configured to:
classify the medical image into one of preset modalities, each preset modality being associated with a different imaging technology;
adjust parameters for a score regression learning model used to determine the calcium score based on the one of preset modalities into which the medical image is classified;
apply a calcium detection model to detect at least one calcium region relevant in determining a calcium score from the medical image;
apply the score regression learning model to the at least one calcium region to determine a calcium score for the medical image; and
provide the determined calcium score of the medical image for a diagnosis of the subject,
wherein the calcium detection model and the score regression learning model are jointly trained by using output information from the calcium detection model as input information into the score regression learning model for determining calcium scores.

2. The system of claim 1 , wherein, to detect at least one calcium region, the at least one processor is further configured to:
detect a region of interest (ROI) containing a target vessel in the medical image; and
detect each calcium region for the medical image by detecting a calcium-containing region in the ROI containing the target vessel.

3. The system of claim 1 , wherein the calcium detection model includes one or more of an active shape model, adaptive thresholding, support vector machine (SVM) classifier, or deep neural network-based model, to identify the at least one calcium region from the medical image.

4. The system of claim 1 , wherein, to detect the at least one calcium region, the at least one processor is further configured to:
determine a bounding box for each calcium region,
wherein, to determine the calcium score for the medical image, the at least one processor is further configured to:
determine a regional calcium score for each bounding box; and
determine the calcium score for the medical image based on the regional calcium score for the at least one calcium region.

5. The system of claim 1 , wherein, to detect the at least one calcium region, the at least one processor is further configured to:
detect a voxel-wise calcium mask from the medical image, wherein the voxel-wise calcium mask including a plurality of voxels,
wherein, to determine the calcium score for the medical image, the at least one processor is further configured to:
determine a regional calcium score for each voxel; and
determine the calcium score for the medical image by summing up the regional calcium scores of the respective voxels included in the voxel-wise calcium mask.

6. The system of claim 1 , wherein, to detect the at least one calcium region, the at least one processor is further configured to:
detect a plurality of calcium regions from the medical image,
wherein the score regression learning model is a recursive neural network, wherein, to determine the calcium score for the medical image, the at least one processor is further configured to:
apply the recursive neural network to the plurality of calcium regions collectively to regress the calcium score in the whole medical image.

7. The system of claim 1 , wherein the score regression learning model is trained by determining sample regional calcium scores based on a ground truth calcium score in training data and distributing the sample regional calcium scores to corresponding ground truth calcium regions derived from the training data.

8. The system of claim 1 , wherein the calcium detection model or the score regression learning model is trained using training data including non-contrast computed tomography (NCCT) images or virtual NCCT images and their corresponding ground truth calcium scores.

9. A computer-implemented method for determining a calcium score from a medical image, comprising:
applying a modality recognition model to classify the medical image into one of preset modalities, each preset modality being associated with a different imaging technology;
adjusting parameters for a score regression learning model used to determine the calcium score based on the one of the present modalities into which the medical image is classified;
receiving the medical image acquired of a subject by an image acquisition device;
applying a calcium detection model to detect at least one calcium region relevant in determining a calcium score from the medical image;
applying the score regression learning model to the at least one calcium region to determine a calcium score for the medical image; and
providing the determined calcium score of the medical image for a diagnosis of the subject,
wherein the calcium detection model and the score regression learning model are jointly trained by using outputting information from the calcium detection model as input information into the score regression learning model for determining calcium scores.

10. The computer-implemented method of claim 9 , wherein detecting at least one calcium region relevant from the medical image further comprises:
applying a region of interest (ROI) module to detect a region of interest (ROI) containing a target vessel in the medical image; and
detecting each calcium region for the medical image by detecting a calcium-containing region in the ROI containing the target vessel.

11. The computer-implemented method of claim 9 , wherein the calcium detection model includes one or more of an active shape model, adaptive thresholding, support vector machine (SVM) classifier, or deep neural network-based model, to identify the at least one calcium region from the medical image.

12. The computer-implemented method of claim 11 , wherein
determining the calcium score for the calcium region further comprises:
determining a bounding box for the calcium region; and

determining the calcium score for the medical image further comprises:
determining a regional calcium score for each bounding box; and
determining the calcium score for the medical image based on the regional calcium score for the at least one calcium region.

13. The computer-implemented method of claim 11 , wherein
detecting the at least one calcium region further comprises:
detecting a voxel-wise calcium mask from the medical image, wherein the voxel-wise calcium mask including a plurality of voxels; and

determining the calcium score for the medical image further comprises:
determining a regional calcium score for each voxel; and
determining the calcium score for the medical image by summing up the regional calcium scores of the respective voxels included in the voxel-wise calcium mask.

14. The computer-implemented method of claim 9 , wherein
detecting the at least one calcium region further comprises:
detecting a plurality of calcium regions from the medical image; and

the score regression learning model is a recursive neural network, and determining the calcium score for the medical image further comprises:
applying a recursive neural network to the plurality of calcium regions collectively to regress the calcium score in the whole medical image.

15. The computer-implemented method of claim 9 , wherein the score regression learning model is trained by determining sample regional calcium scores based on a ground truth calcium score in training data and distributing the sample regional calcium scores to corresponding ground trut

Assignees

• Shenzhen Keya Medical Tech Corporation

Inventors

• Yang Hao-Yu
• Bai Junjie
• Yin Youbing
• Song Qi

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G06T2207/30008

  Bone · CPC title

• G06T2210/41

  Medical · CPC title