CN112070778A - Multi-parameter extraction method based on intravascular OCT and ultrasound image fusion - Google Patents

Abstract

A multi-parameter extraction algorithm based on intravascular OCT and ultrasound image fusion, which is characterized by using the IVUS image and OCT image of the same blood vessel to perform image fusion, and using the fused image to perform deep learning to identify the media of the OCT image. , train a satisfactory artificial intelligence algorithm, and input the OCT image into the artificial intelligence algorithm to segment the media and lumen boundary, so as to obtain the plaque load information. The invention makes up for the defect of insufficient penetrating power of OCT imaging, and can quickly obtain the image load.

Description

A multi-parameter extraction method based on intravascular OCT and ultrasound image fusion

technical field

The invention relates to an image processing technology, in particular to an OCT image processing technology, in particular to a multi-parameter extraction algorithm based on intravascular OCT and ultrasound image fusion.

Background technique

At present, technical means such as X-ray, ultrasound and OCT have been widely used in the diagnosis and treatment of various diseases. The resolution of X-ray is low, but it can image in a large area and overall. It is the basis of current angiographic imaging and the gold standard for coronary artery evaluation and treatment. However, angiography can only provide two-dimensional imaging information, not intraluminal imaging. Tissue composition information; the resolution of intravascular ultrasound is about 100-200 microns, and the penetration depth can reach 10 mm; the resolution of intravascular OCT is about 10-20 microns, and the penetration depth is about 1-2 mm. Currently, these three imaging techniques complement each other clinically and are used in combination. Since the imaging mechanisms of intravascular ultrasound and intravascular OCT are similar, the main difference is resolution and penetration depth. In clinical applications, the main advantage of intravascular ultrasound over intravascular OCT is that it can assess plaque load, but ultrasound signals cannot Penetration of calcified plaque and low-resolution ultrasound signals tend to overestimate lumen area when calculating lumen area. Intravascular OCT has ultra-high resolution and can identify various plaques. Due to the limitation of penetration depth, it cannot effectively identify the location of the media of the plaque, so it cannot provide plaque load information, nor can it be used for lipid plaques. block for efficient imaging.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the problem that the penetrating power of OCT imaging technology is low and cannot effectively identify the location of the medial membrane of the plaque, so it cannot provide plaque load information and cannot effectively image lipid plaques. A multi-parameter extraction method based on intravascular OCT and ultrasound image fusion.

The technical scheme of the present invention is:

A multi-parameter extraction method based on intravascular OCT and ultrasound image fusion, which is characterized by using the IVUS image and OCT image of the same segment of blood vessel to perform image fusion, and using the fused image to perform deep learning to identify the media of the OCT image. After training a satisfactory artificial intelligence algorithm, the OCT image can be input into the artificial intelligence algorithm to segment the media and the lumen boundary, so that the plaque load information can be obtained.

Plaque burden evaluates the mass effect of plaque on the original vascular lumen, and the calculation formula is:

The cross-sectional area of the lamina is usually characterized by the cross-sectional area of the lamina. Due to the large penetration depth of intravascular ultrasound, the media can also be imaged at plaque-enriched sites, so plaque area and media area can be measured and plaque burden can be assessed. The penetration depth of intravascular OCT into tissue is only 1 to 2 mm. When the plaque depth is greater than 1 mm or when special plaques (lipid pools) are encountered, the intima cannot be effectively imaged. Therefore, conventional OCT images cannot evaluate plaque burden. Happening.

The beneficial effects of the present invention are:

The invention makes up for the defect of insufficient penetrating power of OCT imaging, and can quickly obtain the image load.

Description of drawings

FIG. 1 is a schematic flow chart of the algorithm of the present invention.

FIG. 2 is a schematic diagram of the image fusion process of the present invention.

Figure 3 is a schematic diagram showing that the media (green) and lumen (blue) can be segmented by inputting the OCT image into the model, and plaque load information can be obtained.

Figure 4 shows the identification of a section of blood vessels to obtain the location of the lumen and media (arrows) at the plaque. Hemodynamic analysis can be performed to further evaluate the vascular ischemia and local eddy current and stress distribution.

FIG. 5 is a schematic diagram of the U-net network structure.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

As shown in Figure 1-5.

A multi-parameter extraction method based on intravascular OCT and ultrasound image fusion, which uses IVUS pictures and OCT pictures of the same blood vessel to perform image fusion, and uses the fused pictures to perform deep learning to identify the media of the OCT image, and the training results are satisfactory. The artificial intelligence algorithm of the OCT image can be input into the artificial intelligence algorithm to segment the media and lumen boundary, so that the plaque load information can be obtained. Plaque burden evaluates the mass effect of plaque on the original vascular lumen, and the calculation formula is:

The cross-sectional area of the lamina is usually characterized by the cross-sectional area of the lamina. Due to the large penetration depth of intravascular ultrasound, the media can also be imaged at plaque-enriched sites, so plaque area and media area can be measured and plaque burden can be assessed. The penetration depth of intravascular OCT into tissue is only 1 to 2 mm. When the plaque depth is greater than 1 mm or when special plaques (lipid pools) are encountered, the intima cannot be effectively imaged. Therefore, conventional OCT images cannot evaluate plaque burden. Happening.

The specific algorithm flow is shown in Figure 1.

The image fusion process is shown in Figure 2: the IVUS image is input and the media is identified, and the identified media is accurately marked on the OCT image.

The OCT images were then input into the model to segment the media (green) and lumen (blue) and obtain plaque burden information. As shown in Figure 3.

The algorithm can not only obtain the plaque load information, but also identify a section of blood vessels, obtain the media and lumen information of a section of blood vessels, and then calculate the hemodynamics (Fig. 4), identify a section of blood vessels, and obtain plaques The lumen and medial locations at the block (arrows) allow for hemodynamic analysis to further assess vascular ischemia and local eddy and stress distributions.

Network Model: U-net

The U-net network model is one of the most widely used models in the field of semantic segmentation, especially in the medical field. Considering the limited dataset of medical images, we adopted the U-net network model. The biggest advantage of this model is that learning through a smaller training set can also achieve good results.

The U-net network is very simple. The first half is for feature extraction, and the second half is for upsampling. This structure can also be called encoder-decoder. Because the overall structure of the network is similar to the capital letter U, it is called U-net. As shown in Figure 5, Figure 5 is the U-net network semantic segmentation model, the left side down means downsampling, and the right side up means upsampling. The first layer is the input layer (input), the size of the input image is 572*572, through two layers of convolution (the number of volumes (conv) is 64, the convolution size is 3*3, the activation function is ReLU), the maximum pooling After the (max*pool) result is used as the input of one layer, the convolution is continued. After four large convolution pools are used, the downsampling is complete. For the fourth result, deconvolution up-sampling (up-conv) is used, and the step size is 2 to merge the features with the corresponding down-sampling layer. After four downsampling, the output layer (output) adopts a 1*1 convolutional layer and uses the softmax activation function to output the classification prediction result. In the case of limited datasets, data can be enhanced by copying, cropping, and rotating images.

The parts not involved in the present invention are the same as the prior art or can be implemented by using the prior art.

Claims (2)

1. A multi-parameter extraction method based on intravascular OCT and ultrasound image fusion is characterized in that an IVUS picture and an OCT picture of the same section of blood vessel are used for image fusion, deep learning is carried out on the fused picture to identify the tunica media of an OCT image, a satisfactory artificial intelligence algorithm is trained, the OCT picture is input into the artificial intelligence algorithm, the tunica media and a lumen boundary can be segmented, and therefore plaque load information can be obtained.

2. The method of claim 1, wherein the plaque burden assessment is a plaque occupancy effect on the lumen of the native vessel, and is calculated by the formula:

Images