CN116342518A - A Plaque Recognition Method Based on Coronary Artery CT Images - Google Patents

Abstract

The invention discloses a plaque identification method based on a coronary artery CT image, which comprises the steps of firstly preparing a data set, labeling a coronary artery central line, reconstructing a coronary artery cross-sectional image according to the central line, and manually drawing the outline of an inner membrane and an outer membrane of the coronary artery on the cross-sectional image. Then constructing and training a convolutional neural network based on the prepared data set, obtaining a predicted result of intersection points of rays and inner and outer films of the coronary artery after inputting the cross section image of the coronary artery by using the convolutional neural network, connecting the points in sequence, obtaining the outline of the inner and outer films of the coronary artery, and further identifying whether plaque exists on the cross section by calculating the area difference of the inner and outer films. The method can better realize the identification of the outline and plaque of the inner and outer membranes of the coronary artery, and obviously improves the accuracy and repeatability. The plaque identification time is shortened, the efficiency is obviously improved, and the method has higher auxiliary value for diagnosis of clinicians.

Description

A Plaque Recognition Method Based on Coronary Artery CT Images

technical field

The invention relates to the technical field of image recognition, in particular to a plaque recognition method based on coronary CT images.

Background technique

Coronary artery CT is central to the diagnosis and risk stratification of coronary artery disease. From the quantification of coronary artery calcification by Agaston score, to the risk classification of luminal stenosis, to the description of plaque composition, the qualitative characteristics of high-risk plaques, and the analysis of plaque radiomics parameters, the role of coronary artery CT in plaque assessment has always been in progress. In current clinical practice, the plaque recognition algorithm mainly relies on the gradient feature of the gray value of the image. The coronary artery contour identified by it has a large error, and usually requires a lot of manual optimization to achieve the doctor's satisfactory result. This process takes a long time. and is susceptible to inter-observer variability. With the development of deep learning technology, the intelligent analysis of medical images has ushered in great progress, which provides new possibilities for plaque recognition tasks.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention aims to provide a plaque identification method based on coronary artery CT images.

In order to achieve the above object, the present invention adopts the following technical solutions:

A plaque identification method based on coronary artery CT images, the specific process is:

S1. Select a set number of coronary artery CT images, manually mark the centerlines of the three main coronary artery branches in each coronary artery CT image, and use the marked centerlines to perform surface reconstruction for each vessel in the image. Coronary artery cross-sectional images; and then in each coronary artery cross-sectional image, the contours of the coronary artery intima and intima were manually marked;

The cross-sectional images of the coronary arteries manually marked with the contours of the inner and outer membranes of the coronary arteries were randomly divided into a training set and a test set according to the set ratio;

For each cross-sectional image of the coronary artery, the center of the image is used as the origin to emit rays in all directions, and the distance from the intersection point of the ray with the inner and outer layers of the coronary artery to the origin is used as the gold standard for model prediction; the cross-sectional image of the coronary artery is input to the convolution The neural network model, through multi-layer feature extraction, outputs the prediction results of the distance from the intersection point of the ray at each angle and the coronary artery intima to the origin; for the training set image, the prediction result of the image by the convolutional neural network is compared with the gold standard The difference between the network parameters is adjusted by the stochastic gradient descent algorithm to achieve the purpose of reducing the prediction error of the convolutional neural network model; for the test set image, the convolutional neural network is used to measure the difference between the prediction result of the test set image and the gold standard. performance of neural network models;

S2. Use the convolutional neural network model obtained through training to predict the intersection points of the rays emitted in all directions from the center of the image and the coronary artery intima to the input coronary artery cross-sectional image, and sequentially connect the intersection points to obtain the coronary artery intima and intima forecast result;

S3. Plaque identification: connect the intersection of the ray predicted by the convolutional neural network model and the intima and intima of the coronary artery to obtain the contour of the coronary artery, and select the cutoff with an area difference of intima and intima greater than 0.1 ^mm2 and a length greater than 2 mm along the center line as the plaque Block prediction results.

Further, in step S2, the set ratio is 8:2.

The present invention also provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and the above-mentioned method is realized when the computer program is executed by a processor.

The present invention also provides a computer device, including a processor and a memory, and the memory is used to store a computer program; when the processor is used to execute the computer program, the above method is realized.

The beneficial effect of the present invention is that: the method of the present invention can better realize the delineation of the inner and outer membranes of the coronary arteries and the recognition of the plaques, and the accuracy and repeatability are obviously improved. The plaque identification time is shortened, the efficiency is significantly improved, and it has a high auxiliary value for the diagnosis of clinicians.

Description of drawings

Figure 1 is a schematic diagram of the implementation of step S1.1 in the embodiment of the present invention;

Fig. 2 is a schematic diagram of the implementation of step S1.3 in the embodiment of the present invention;

Fig. 3 is a schematic diagram of the implementation of step S3 in the embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. It should be noted that this embodiment is based on the technical solution, and provides detailed implementation and specific operation process, but the protection scope of the present invention is not limited to the present invention. Example.

This embodiment provides a plaque recognition method based on coronary artery CT images, the specific process is:

S1. As shown in Figure 1, select 100 cases of coronary artery CT images with high image quality, manually mark the centerlines of the three main coronary artery branches in each case of coronary artery CT images, and use the annotation method for each blood vessel in the image A good centerline was reconstructed to generate a cross-sectional image of the coronary artery; then the contours of the inner and outer layers of the coronary artery were manually marked in each cross-sectional image of the coronary artery.

It should be noted that the three main coronary artery branches include the left anterior descending coronary artery, the left circumflex coronary artery and the right coronary artery.

100 cases of coronary artery cross-sectional images manually marked with the contour of coronary artery intima and intima were randomly divided into training set and test set according to the ratio of 8:2.

As shown in Figure 2, for each coronary artery cross-sectional image, rays are emitted in all directions with the center of the image as the origin, and the distance from the intersection point of the ray with the inner and outer layers of the coronary artery to the origin is used as the gold standard for model prediction; the coronary artery is transected The surface image is input to the convolutional neural network model, and the prediction results of the distance from the intersection point of the ray at each angle to the coronary artery intima to the origin are output through multi-layer feature extraction; for the training set images, the training set images are compared by the convolutional neural network. The difference between the predicted results and the gold standard, using the stochastic gradient descent algorithm to adjust the network parameters to achieve the purpose of reducing the prediction error of the convolutional neural network model; for the test set image, using the convolutional neural network to predict the test set image and Gold standard difference to measure the performance of convolutional neural network models. The average error of the prediction results of the coronary artery intima and intima on the test set obtained by the model constructed in this embodiment is less than 0.52mm, and the average processing time per case is 47.2 seconds; the average error of two independent observers manually marking the results in the test set is 0.54 mm, the average treatment time per case was 32.5 minutes. Experiments have proved that the performance of the model constructed in this embodiment is better than the inter-observer differences in the test set.

S2. Use the convolutional neural network model obtained through training to predict the intersection points of the rays emitted in all directions from the center of the image and the coronary artery intima to the input coronary artery cross-sectional image, and sequentially connect the intersection points to obtain the coronary artery intima and intima forecast result.

S3. Plaque identification: As shown in Figure 3, connect the intersection of the ray predicted by the convolutional neural network model and the intima ^and intima of the coronary artery to obtain the coronary artery contour. The 2mm cutoff is the plaque prediction area. Further parameters such as volume, load, and reconstruction index can be calculated for the plaque area.

For those skilled in the art, various corresponding changes and modifications can be made according to the above technical solutions and concepts, and all these changes and modifications should be included in the protection scope of the claims of the present invention.

Claims (4)

1. A plaque identification method based on coronary artery CT image, is characterized in that, concrete process is: S1. Select a set number of coronary artery CT images, manually mark the centerlines of the three main coronary artery branches in each coronary artery CT image, and use the marked centerlines to perform surface reconstruction for each vessel in the image. Coronary artery cross-sectional images; and then in each coronary artery cross-sectional image, the contours of the coronary artery intima and intima were manually marked; The cross-sectional images of the coronary arteries manually marked with the contours of the inner and outer membranes of the coronary arteries were randomly divided into a training set and a test set according to the set ratio; For each cross-sectional image of the coronary artery, the center of the image is used as the origin to emit rays in all directions, and the distance from the intersection point of the ray with the inner and outer layers of the coronary artery to the origin is used as the gold standard for model prediction; the cross-sectional image of the coronary artery is input to the convolution The neural network model, through multi-layer feature extraction, outputs the prediction results of the distance from the intersection of the ray at each angle and the coronary artery intima to the origin; for the training set images, the prediction results of the images by the convolutional neural network are compared with the gold standard Difference, using the stochastic gradient descent algorithm to adjust the network parameters to achieve the purpose of reducing the prediction error of the convolutional neural network model; performance of the network model; S2. Use the convolutional neural network model obtained through training to predict the intersection points of the rays emitted in all directions from the center of the image and the coronary artery intima to the input coronary artery cross-sectional image, and sequentially connect the intersection points to obtain the coronary artery intima and intima forecast result; S3. Plaque identification: connect the intersection of the ray predicted by the convolutional neural network model and the intima and intima of the coronary artery to obtain the contour of the coronary artery, and select the cutoff with an area difference of intima and intima greater than 0.1 ^mm2 and a length greater than 2 mm along the center line as the plaque Block prediction results. 2. The method according to claim 1, characterized in that, in step S2, the set ratio is 8:2. 3. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method according to any one of claims 1-2 is implemented. 4. A computer device, characterized in that it comprises a processor and a memory, and the memory is used to store a computer program; when the processor is used to execute the computer program, the method according to any one of claims 1-2 is realized .

Images