Intravascular ultrasonic lithotripsy system integrating IVUS and IVL

Abstract

The invention relates to an intravascular ultrasonic lithotripsy system integrating IVUS and IVL. The system comprises a lithotripsy integration unit which at least comprises an imaging tube group and an ultrasonic array element assembly adaptively connected with the imaging tube group, and a lithotripsy control unit which is used for performing calcification characteristic monitoring processing on an intravascular ultrasonic image group of a target blood vessel acquired by an IVUS array element group, the calcification feature level and the corresponding calcification ablation state of the target calcification area are at least determined after calcification feature monitoring processing, and when the IVL array element set is configured to enter the intravascular shock wave lithotripsy work, the lithotripsy control unit at least configures the impact lithotripsy working parameters of the IVL array element set to be matched with the calcification feature level of the target calcification area. According to the invention, effective integration of IVUS and IVL can be effectively realized, and the precision and reliability of intravascular ultrasonic lithotripsy can be improved.

Description

Technical Field

[0001] This invention relates to an ultrasonic lithotripsy system, and more particularly to an intravascular ultrasonic lithotripsy system integrating IVUS and IVL. Background Technology

[0002] Coronary artery calcification is a common pathological change in cardiovascular disease, and it can seriously affect the success rate and long-term prognosis of percutaneous coronary intervention (PCI). Traditional treatments for coronary artery calcification, such as balloon dilation, balloon cutting, and rotational atherectomy, all carry risks such as vascular dissection, perforation, and instrument abrasion. While intravascular ultrasound (IVUS) can clearly visualize calcified lesions, it lacks direct therapeutic capabilities. Intravascular shock wave lithotripsy (IVL), as an emerging technology, can break up calcified plaques using pulsed pressure waves, but currently, IVL surgery largely relies on external imaging guidance, making it impossible to locate and monitor treatment effects in real time and with precision.

[0003] As explained above, most existing IVL techniques are single-function shockwave devices. If combined with IVUS (intravascular ultrasound) imaging, an additional imaging catheter is required, leading to complex procedures and a large space requirement within the vessel, making them unsuitable for small vessels such as coronary arteries. Furthermore, in non-integrated designs, the IVUS and shockwave catheters need to be positioned separately. This can result in a mismatch between the "imaging position" and the "shockwave release position" due to vascular peristalsis or operational errors, affecting the precise fragmentation of calcifications.

[0004] In existing technologies, when using IVUS image guidance, the number of shockwave releases and the energy depend on the surgeon's experience, posing a risk of "overtreatment" or "undertreatment".

[0005] In summary, how to effectively integrate IVUS and IVL, and how to effectively improve the reliability of intravascular ultrasound lithotripsy after integration, are urgent technical challenges that need to be addressed. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an intravascular ultrasound lithotripsy system that integrates IVUS and IVL, which can effectively integrate IVUS and IVL and improve the accuracy and reliability of intravascular ultrasound lithotripsy.

[0007] According to the technical solution provided by the present invention, an intravascular ultrasound lithotripsy system integrating IVUS and IVL is provided, the intravascular ultrasound lithotripsy system comprising:

[0008] The lithotripsy integrated unit includes at least an imaging tube assembly and an ultrasonic array element component adapted and connected to the imaging tube assembly, wherein...

[0009] The ultrasound array assembly includes at least an IVUS array for acquiring intravascular ultrasound images and an IVL array for performing intravascular shock wave lithotripsy. The IVL array and the IVUS array are integrated on the same backing and mounted on the imaging tube assembly through the backing.

[0010] The lithotripsy control unit performs calcification feature monitoring on the intravascular ultrasound images of the target vessel acquired via IVUS array elements. This calcification feature monitoring process determines at least the calcification characteristic level and corresponding calcification ablation status of the target calcified region.

[0011] When the calcification ablation state matches the lithotripsy termination condition, the lithotripsy control unit stops the IVL array group from performing intravascular shock wave lithotripsy; otherwise, the IVL array group is configured to enter intravascular shock wave lithotripsy operation.

[0012] When configuring the IVL array element to perform intravascular shock wave lithotripsy, the lithotripsy control unit shall at least configure the shock wave lithotripsy parameters of the IVL array element to match the calcification characteristic level of the target calcified area.

[0013] The imaging tube assembly includes an inner tube and an integrated outer tube fitted onto the inner tube, wherein...

[0014] The backing sleeve is placed on the inner tube, and the backing sleeve, IVUS array elements, and IVL array elements are covered by a balloon assembled on the inner tube.

[0015] The IVUS elements in the IVUS array group and the IVL elements in the IVL array group are arranged in alternating rings on the backing. The IVUS elements and IVL elements are coplanar on the backing, and the ultrasonic emission direction of the IVL elements is radially aligned with the detection direction of the IVUS elements.

[0016] The integrated outer tube is at least fitted over the tail end of the inner tube, and the balloon is located between the head end of the integrated outer tube and the head end of the inner tube.

[0017] When performing calcification feature monitoring on intravascular ultrasound images of the target blood vessel, the following steps are included:

[0018] Calcification region identification processing is performed sequentially on intravascular ultrasound images within the intravascular ultrasound image group. After calcification region identification processing is performed on all intravascular ultrasound images, the target calcification region, the target feature state of the target calcification region, and the calcification feature level of the target calcification region corresponding to the current intravascular ultrasound image group are determined.

[0019] The target characteristic state of the target calcification region includes at least the target calcification angle and / or the target calcification thickness;

[0020] The calcification characteristic grades include severe calcification, moderate calcification, and / or mild calcification.

[0021] When configuring the impact crushing parameters of the IVL array to match the calcification characteristic level of the identified calcified region, the following are included:

[0022] When the calcification characteristic level of the target calcified area is severe calcification, configure the impact crushing working parameters of the IVL array element to put the IVL array element in high-energy mode.

[0023] When the calcification characteristic level of the target calcified area is moderate calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in medium energy mode.

[0024] When the calcification characteristic level of the target calcified area is light calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in low-energy mode.

[0025] The impact crushing operating parameters of the IVL array element group include at least the array element excitation voltage, excitation pulse width, excitation duty cycle, and energy density.

[0026] The conditions for terminating the impact of the crushed stone include a first condition for termination and / or a second condition for termination, wherein...

[0027] When the calcification ablation state meets either the first or second condition for impact termination, then the calcification characteristic change is matched with the impact termination condition of the crushed stone.

[0028] The first condition for terminating the impact includes an increase in vessel diameter of not less than the diameter increase threshold and a decrease in calcification thickness of not less than the thickness decrease threshold.

[0029] The second condition for terminating the impact includes the interruption of calcification continuity in the target calcified area, an increase in vessel diameter of not less than the diameter increase threshold, and a decrease in plaque echo intensity of not less than the echo intensity decrease threshold.

[0030] When performing calcification region identification processing on intravascular ultrasound images, the following steps are included:

[0031] A calcification segmentation model was constructed and deployed within the gravel control unit.

[0032] When performing calcification region identification processing on intravascular ultrasound images, a calcification segmentation model is used to segment and identify each pixel in the intravascular ultrasound image. After segmentation and identification, the mask type of each pixel is determined, and the intravascular ultrasound image is segmented into basic calcification region and background region based on the mask type of each pixel. The mask type of each pixel is either foreground mask or background mask.

[0033] Based on the basic calcification regions of all intravascular ultrasound images, a target calcification region corresponding to the current intravascular ultrasound image is generated.

[0034] For each basic calcification region in an intravascular ultrasound image, calculate the basic calcification angle and basic calcification thickness corresponding to the basic calcification region.

[0035] Based on all the basic calcification angles, generate the target calcification angle;

[0036] Based on all the basic calcification thicknesses, the target calcification thickness is generated.

[0037] The calcification segmentation model includes a basic segmentation network and a group of acoustic-shadow units adapted and connected to the basic segmentation network, wherein...

[0038] The basic segmentation network includes an encoder network and a decoder network adapted and connected to the encoder network. The encoder network includes several encoding layers, and the decoder network includes several decoding layers. The number of decoding layers in the decoder network is the same as the number of encoding layers in the encoder network, and the encoder network and decoder network are U-shaped after being adapted and connected.

[0039] The sound and shadow unit group includes several sound and shadow units. The number of sound and shadow units is less than the number of coding layers in the encoder network. Each sound and shadow unit acquires the mid-to-deep features generated by a coding layer in the encoder network and performs calcification-sound and shadow association on the acquired mid-to-deep features to generate a sound and shadow target feature map. The sound and shadow target feature map is used to model the directional dependency relationship between the calcified strong echo region and the sound and shadow region behind the calcified strong echo region.

[0040] The sound and shadow target feature map generated by each sound and shadow unit is loaded into the decoding layer that corresponds to the current coding layer to supplement the semantic understanding of the strong echo-low echo association of the decoding layer.

[0041] Each coding layer in the encoder network outputs a coding feature map, wherein the mid-to-deep features are generated by downsampling the coding feature map output by the coding layer located in the mid-to-deep position;

[0042] When the acquired mid-to-deep features are generated by downsampling the encoded feature map of the deepest encoding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is upsampled, and in the corresponding decoding layer, the sound and shadow upsampled feature map and the corresponding encoded feature map are concatenated.

[0043] When the acquired mid-to-deep features are generated by downsampling the coding feature map of the mid-domain coding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is added to the corresponding decoding feature map, and the sound and shadow decoding feature map is generated after the addition.

[0044] The audio-visual decoding feature map is upsampled to generate an upsampled audio-visual decoding feature map. The generated upsampled audio-visual decoding feature map is then concatenated with the corresponding encoded feature map output by the encoding layer. The decoded feature map is the feature map generated by the previous decoding layer.

[0045] The audio-visual unit includes an audio-visual module and an audio-visual post-processing module, wherein...

[0046] The audio-visual module includes a semantic alignment enhancement module and a directional dependency modeling unit connected in sequence, wherein,

[0047] The semantic alignment enhancement module receives the loaded mid-to-deep features and performs semantic alignment enhancement processing on the mid-to-deep features. After semantic alignment enhancement processing, the contrast between the calcified strong echo region and the sound shadow region is focused, and a mid-to-deep semantic alignment enhancement feature map is generated.

[0048] The spatial orientation capture process of the mid-to-deep semantic alignment enhancement feature map is performed by the directional dependency modeling unit to capture the directional propagation features of the calcified strong echo region and the sound shadow region, and to generate the sound shadow basic feature map. The feature dimension of the sound shadow basic feature map is consistent with the feature dimension of the mid-to-deep features.

[0049] The audio-visual post-processing module performs at least feature processing on the basic audio-visual feature map to generate the audio-visual target feature map after feature processing.

[0050] The directional dependency modeling unit may include a difference mapping module, a direction-aware module, and a radial spatial attention module connected in sequence, wherein...

[0051] The difference mapping module performs feature filtering and difference mapping on the mid-to-deep semantic alignment enhancement feature map, and generates a strong echo-low echo difference feature map. The feature filtering and difference mapping process includes two branches. In one branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a sigmoid function operation. In the other branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a tanh operation. After that, the results of the two branches are multiplied point by point.

[0052] After generating the strong echo-low echo difference feature map, the direction perception module performs direction perception processing on the strong echo-low echo difference feature map so that a direction perception feature map can be generated after the direction perception processing. In the direction perception processing, at least a 3×3 convolution with angle weights is used to perform convolution processing on the strong echo-low echo difference feature map.

[0053] After generating the orientation-aware feature map, the radial spatial attention module first performs radial spatial attention processing to generate a radial spatial attention feature map. Then, the radial spatial attention feature map and the orientation-aware feature map are multiplied point by point to generate the basic sound and shadow feature map after the point-by-point multiplication.

[0054] The advantages of this invention are: integrating the IVUS array elements in the IVUS array group and the IVL array elements in the IVL array group onto the same backing, and assembling the backing onto the inner tube to form a lithotripsy integrated unit, thereby effectively integrating IVUS and IVL; thus, in a single percutaneous coronary intervention, intravascular ultrasound imaging and intravascular shock wave lithotripsy can be achieved, improving the efficiency and reliability of intravascular ultrasound lithotripsy.

[0055] After intravascular shock wave lithotripsy is performed on the target calcified area of ​​the target blood vessel using the IVL array, the lithotripsy control unit determines the corresponding calcification ablation state and compares the calcification ablation state with the lithotripsy termination condition to determine the working state of the subsequent IVUS array and IVL array. This can avoid excessive or insufficient shock wave lithotripsy on the target calcified area, thereby improving the accuracy and reliability of intravascular ultrasound lithotripsy.

[0056] When performing calcification region identification processing on intravascular ultrasound images, a calcification segmentation model is used for segmentation and identification. The deep features in the acoustic shadow unit group within the calcification segmentation model are used to perform calcification-acoustic shadow correlation, so as to use the acoustic shadow behind the calcification region for auxiliary segmentation and optimize the segmentation accuracy. This can improve the accuracy and reliability of segmentation and identification. Attached Figure Description

[0057] Figure 1 This is a structural block diagram of one embodiment of the intravascular ultrasonic lithotripsy system of the present invention.

[0058] Figure 2 This is a perspective view of one embodiment of the crushed stone integration unit of the present invention.

[0059] Figure 3 To be Figure 2 A schematic diagram of one embodiment after the outer tube of the integrated circuit is removed.

[0060] Figure 4 To be Figure 2 A schematic diagram of one embodiment of the integrated outer tube and balloon removal.

[0061] Figure 5 To be Figure 4 A schematic diagram of one embodiment after the inner tube has been removed.

[0062] Figure 6 This is a schematic diagram of one embodiment of the calcification segmentation model of the present invention.

[0063] Figure 7 This is a schematic diagram of one embodiment of the audio-visual module of the present invention.

[0064] Explanation of reference numerals in the attached diagram: 1-inner tube, 2-balloon, 3-integrated outer tube, 4-IVL element cable, 5-IVUS element cable, 6-backing seat, 7-backing, 8-IVL element, 9-IVUS element, 100-integrated lithotripsy unit, and 200-lithotripsy control unit. Detailed Implementation

[0065] The present invention will be further described below with reference to specific accompanying drawings and embodiments.

[0066] To effectively integrate IVUS and IVL and improve the accuracy and reliability of intravascular ultrasound lithotripsy, this invention provides an intravascular ultrasound lithotripsy system integrating IVUS and IVL. Specifically, the intravascular ultrasound lithotripsy system includes:

[0067] The lithotripsy integration unit 100 includes at least an imaging tube assembly and an ultrasonic array element component adapted and connected to the imaging tube assembly, wherein...

[0068] The ultrasound array assembly includes at least an IVUS array for acquiring intravascular ultrasound images and an IVL array for performing intravascular shock wave lithotripsy. The IVL array and the IVUS array are integrated on the same backing 7 and mounted on the imaging tube assembly through the backing 7.

[0069] The lithotripsy control unit 200 performs calcification feature monitoring processing on the intravascular ultrasound images of the target vessel acquired via IVUS array elements. This calcification feature monitoring processing determines at least the calcification feature level and the corresponding calcification ablation status of the target calcified region.

[0070] When the calcification ablation state matches the lithotripsy termination condition, the lithotripsy control unit 200 stops the IVL array group from performing intravascular shock wave lithotripsy; otherwise, the IVL array group is configured to enter intravascular shock wave lithotripsy operation.

[0071] When configuring the IVL array element to perform intravascular shock wave lithotripsy, the lithotripsy control unit 200 configures the shock wave lithotripsy parameters of the IVL array element to match the calcification characteristic level of the target calcified area.

[0072] It should be noted that the intravascular ultrasound lithotripsy system of the present invention integrates IVUS and IVL, thus enabling intravascular ultrasound imaging and intravascular shock wave lithotripsy to be performed in a single percutaneous coronary intervention. Figure 1The figure illustrates an embodiment of the intravascular ultrasound lithotripsy system of the present invention. As shown in the figure, the intravascular ultrasound lithotripsy system includes a lithotripsy integration unit 100 and a lithotripsy control unit 200. The lithotripsy integration unit 100 integrates IVUS and IVL. The lithotripsy control unit 200 is adapted to and connected to the lithotripsy integration unit 100 so that the working state of the lithotripsy integration unit 100 can be controlled by the lithotripsy control unit 200. For example, the lithotripsy integration unit 100 can be controlled to be in IVUS imaging state or intravascular shock wave lithotripsy state. IVUS imaging is intravascular ultrasound imaging. The method and process by which the lithotripsy control unit 200 controls the working state of the lithotripsy integration unit 100 will be described in detail below.

[0073] To meet the needs of intravascular ultrasound imaging and intravascular shock wave lithotripsy, in one embodiment of the present invention, the lithotripsy integration unit 100 should include an imaging tube assembly and an ultrasound array component. The imaging tube assembly enables percutaneous coronary intervention, and the ultrasound array component is used for intravascular ultrasound imaging and intravascular shock wave lithotripsy. Specifically, the ultrasound array component should include an IVUS array assembly and an IVL array assembly. To achieve effective integration of IVUS and IVL, in one embodiment of the present invention, the IVUS array assembly and the IVL array assembly are integrated on the same backing 7, such as... Figure 5 As shown, it can also be mounted on the imaging tube assembly via the backing 7.

[0074] In one embodiment of the present invention, the imaging tube assembly includes an inner tube 1 and an integrated outer tube 3 sleeved on the inner tube 1, wherein,

[0075] The backing 7 is placed on the inner tube 1, and the backing 7, IVUS array elements and IVL array elements are covered by the balloon 2 assembled on the inner tube 1.

[0076] IVUS element 9 in the IVUS array group and IVL element 8 in the IVL array group are arranged in an alternating ring on the backing 7. IVUS element 9 and IVL element 8 are coplanar on the backing 7, and the ultrasonic emission direction of IVL element 8 is radially aligned with the detection direction of IVUS element 9.

[0077] The integrated outer tube 3 is at least fitted onto the tail end of the inner tube 1, and the balloon 2 is located between the head end of the integrated outer tube 3 and the inner tube 1.

[0078] Figure 1 The image shows one embodiment of the crushed stone integration unit 100, which consists of... Figures 1-4As can be seen, the imaging tube assembly may include an inner tube 1 and an integrated outer tube 3. The inner tube 1 is hollow, and generally, the lumen diameter of the inner tube 1 can be 0.36mm to 0.46mm to accommodate at least a 0.014-inch guidewire, ensuring smooth passage of the guidewire and providing basic support for the precise positioning and advancement of the imaging tube assembly within the blood vessel. The inner tube 1 can be made of commonly used materials, specifically those that can meet the requirements for positioning within the blood vessel.

[0079] Figure 1 In this design, an integrated outer tube 3 is fitted onto an inner tube 1, and the outer tube 3 and inner tube 1 are coaxially distributed. Generally, the outer diameter of the catheter formed between the outer tube 3 and the inner tube 1 is ≤1.2mm. Furthermore, a balloon 2 is also installed on the inner tube 1, typically located at the tip of the inner tube 1. Figure 1 The image shows an embodiment where the balloon 2 is in a fluid-filled, inflated state. Figure 1 In this configuration, the integrated outer tube 3 is fitted onto the tail of the inner tube 1, and one end of the integrated outer tube 3 adjacent to the head of the inner tube 1 is fixedly connected to the balloon 2. Furthermore, the balloon 2 can be made of Pebax or nylon material, with a hydrophilic coating (such as polyvinylpyrrolidone) on its surface to reduce the coefficient of friction to ≤0.1, thereby minimizing damage to the blood vessel wall and improving the delivery performance of the imaging tube assembly within the blood vessel.

[0080] Figure 4 The diagram shows a schematic of an embodiment in which the backing 7 is fitted onto the inner tube 1. As illustrated, in the above description, the IVL array element group and the IVUS array element group are assembled onto the imaging tube assembly via the backing 7. Specifically, the IVL array element group and the IVUS array element group are integrated onto the same backing 7 and assembled onto the inner tube 1 via the backing 7, thereby achieving a compatible connection between the ultrasound array element assembly and the imaging tube assembly. Furthermore... Figure 1 In the illustrated embodiment, after the backing 7 is assembled onto the inner tube 1, the backing 7, the IVUS array elements, and the IVL array elements are covered by the balloon 2 assembled onto the inner tube 1. Specifically, the length of the balloon 2 on the inner tube 1 is greater than the length of the backing 7, that is, the backing 7, the IVUS array elements, and the IVL array elements can be encapsulated on the inner tube 1 by the balloon 2.

[0081] It should be noted that balloon 2 is generally filled with fluid (usually saline). This fluid serves as a conduit for the shock waves emitted by the IVL array during operation, efficiently delivering the shock waves to the calcified plaques on the vessel wall. The shock waves selectively fragment the hard calcified tissue while causing minimal damage to the surrounding normal vessel wall (such as elastic tissue and muscle layer), achieving "targeted lithotripsy." Furthermore, during operation, balloon 2 maintains contact with the vessel wall after inflation, ensuring that the shock wave energy is more concentrated on the calcified area, avoiding energy dispersion and improving lithotripsy efficiency. Moderate inflation of balloon 2 can also slightly dilate the vessel while simultaneously fixing the position of the imaging tube array, preventing displacement of the lithotripsy unit 100 during treatment and ensuring the precision of the shock wave action. Moreover, compared to traditional high-pressure balloon dilation (which may tear blood vessels), this method, through fluid-conducted shock waves, can achieve calcified fragmentation at lower pressure, reducing the risk of complications such as vascular dissection and perforation.

[0082] To effectively acquire intravascular ultrasound images, the IVUS array group should include multiple IVUS elements 9; furthermore, to effectively achieve intravascular shock wave lithotripsy, the IVL array group should include multiple IVL elements 8. The IVUS array group and IVL array group are integrated on the backing 7, specifically meaning that all IVUS elements 9 and all IVL elements 8 are simultaneously integrated on the same backing 7. In one embodiment of the present invention, the IVUS elements 9 and IVL elements 8 are arranged in an alternating ring pattern on the backing 7. Figure 4 and Figure 5 An embodiment of IVUS array element 9 and IVL array element 8 arranged on backing 7 is shown.

[0083] To fit snugly onto the inner tube 1, the backing 7 can be a hollow cylinder. IVUS elements 9 and IVL elements 8 are mounted on the outer surface of the backing 7, arranged alternately. For example, for one IVUS element 9, one IVL element 8 is distributed on each side of the IVUS element 9, and vice versa. Furthermore, multiple IVUS elements 9 are evenly distributed on the backing 7, and similarly, multiple IVL elements 8 are also evenly distributed on the backing 7. The number of IVUS elements 9 and IVL elements 8 can be selected as needed to achieve the desired alternating and uniform arrangement, and to meet the requirements for acquiring intravascular ultrasound images and intravascular shock wave lithotripsy.

[0084] In practical implementation, IVUS array element 9 and IVL array element 8 can adopt commonly used array element forms, specifically designed to enable the formation of an IVUS array element group using all IVUS array elements 9, and to acquire intravascular ultrasound images of the target blood vessel using the IVUS array element group. Similarly, IVL array element 8 should be designed to enable intravascular shock wave lithotripsy after the formation of the IVL array element group. IVUS array elements 9 and IVL array elements 8 can be integrated onto the backing 7 using techniques commonly used in this field, such as mounting IVUS array elements 9 and IVL array elements 8 onto the backing 7. Specific integration methods onto the backing 7 will not be elaborated here.

[0085] When IVUS element 9 and IVL element 8 are attached to the outer surface of backing 7, IVUS element 9 and adjacent IVL element 8 are coplanar. In addition, the ultrasound emission direction of IVL element 8 is radially aligned with the detection direction of IVUS element 9. This allows for the acquisition of intravascular ultrasound images using the IVUS element group, and after identifying the target calcified area, the IVL element group can be used to perform intravascular shock wave lithotripsy on the target calcified area, thereby improving the accuracy and reliability of intravascular shock wave lithotripsy.

[0086] To mount the backing 7 onto the inner tube 1, the backing 7 can first be mounted onto the backing seat 6, and then the backing 7 can be fixed onto the inner tube 1 via the backing seat 6. Furthermore, IVL element cables 4 and IVUS element cables 5 can be mounted on the backing 7. Each IVL element cable 4 can be electrically connected to one IVL element 8, and each IVUS element cable 5 can be electrically connected to one IVUS element 9. The IVL element cables 4 and IVUS element cables 5 can be used to drive the IVL element 8 and IVUS element 9 respectively. The specific method of driving the IVL element 8 and IVUS element 9 can be consistent with existing technology, aiming to meet the requirements of intravascular shock wave lithotripsy and the acquisition of intravascular ultrasound images. Figure 1 and Figure 2 It can be seen that IVL array element cable 4 and IVUS array element cable 5 should be located inside the integrated outer tube 3.

[0087] When the lithotripsy integration unit 100 adopts the above-described form, the connection between the lithotripsy integration unit 100 and the lithotripsy control unit 200 specifically means that at least the IVL array group and the IVUS array group are electrically connected to the lithotripsy control unit 200, so that the lithotripsy control unit 200 can control the working state of the IVL array group and the IVUS array group. In one embodiment of the present invention, the lithotripsy control unit 200 is configured to operate the IVL array group and the IVUS array group simultaneously. Specifically, when the IVL array group is used to emit shock waves, the operation of the IVUS array group is paused; when the IVUS array group is used to acquire intravascular ultrasound imaging, the emission of shock waves using the IVL array group is paused to avoid mutual interference and ensure the stable operation of their respective functions.

[0088] It should be understood that the lithotripsy control unit 200 can be a control device that can control the simultaneous operation of the IVL array element group and the IVUS array element group. Of course, the lithotripsy control unit 200 should also be able to perform calcification feature monitoring and processing on the intravascular ultrasound images of the target blood vessel. The target blood vessel specifically refers to the blood vessel that needs to acquire intravascular ultrasound images and undergo intravascular shock wave lithotripsy. The type of lithotripsy control unit 200 can be selected as needed, which will not be elaborated here.

[0089] In practical implementation, when configuring an IVUS array, the imaging frame rate of the IVUS array can be ≥60 frames / second. Therefore, when performing one IVUS imaging session, multiple intravascular ultrasound images can be acquired. These multiple intravascular ultrasound images can form an intravascular ultrasound image group. It can be understood that the intravascular ultrasound images within the intravascular ultrasound image group have temporal characteristics. Furthermore, when configuring an IVL array for one intravascular shock wave lithotripsy session, the corresponding time can be 10ms-100ms.

[0090] After the lithotripsy control unit 200 of this invention performs calcification feature monitoring processing on the intravascular ultrasound image group, it can determine the target calcification area corresponding to the intravascular ultrasound image group, the calcification feature level of the target calcification area, and the calcification ablation state of the target calcification area. Specifically, the target calcification area is the calcification area in the target blood vessel to be subjected to shock wave lithotripsy; the calcification ablation state of the target calcification area specifically refers to the calcification change state of the target calcification area; the calcification feature level of the target calcification area specifically refers to the degree of calcification of the target calcification area, that is, the grading of the degree of calcification of the target calcification area.

[0091] To avoid overtreatment or undertreatment of the target calcified area, after determining the calcification ablation status of the target calcified area, the calcification ablation status should be compared with the lithotripsy termination conditions. Specifically, when the calcification ablation status matches the lithotripsy termination conditions, it indicates that the target calcified area does not require further intravascular shock wave lithotripsy. In this case, the lithotripsy control unit 200 should stop the IVL array from performing intravascular shock wave lithotripsy, and the lithotripsy control unit 200 should put the IVL array into a stopped working state. However, if the calcification ablation status does not match the lithotripsy termination conditions, in order to avoid undertreatment, the lithotripsy control unit 200 should control the IVL array to enter intravascular shock wave lithotripsy to perform shock wave lithotripsy on the target calcified area within the target blood vessel.

[0092] In addition, to further improve the accuracy and reliability of shock wave lithotripsy, when the IVL array is configured to operate in the blood vessel for shock wave lithotripsy, the lithotripsy control unit 200 configures the shock wave lithotripsy operating parameters of the IVL array to match the calcification characteristic level of the target calcified area. That is, when performing intravascular shock wave lithotripsy on the target calcified area, overtreatment or undertreatment can be avoided, which can improve the reliability of shock wave lithotripsy and also improve the efficiency of shock wave lithotripsy.

[0093] It should be understood that when the lithotripsy integration unit 100 adopts the above-described structural form, endovascular intervention of the lithotripsy integration unit 100 can be achieved using existing technical means. After endovascular intervention, the intravascular ultrasound image obtained using the IVUS array should include the target calcified area. After determining the target calcified area of ​​the target blood vessel, the IVL array can be used to perform intravascular shock wave lithotripsy on the target calcified area. That is, the present invention only changes the integration method of the IVUS array and the IVL array; the specific working methods of intravascular ultrasound imaging and intravascular shock wave lithotripsy can be consistent with existing technologies. In operation, the intravascular ultrasound image set can generally be obtained first using the IVUS array, and then the determination of whether to use the IVL array for intravascular shock wave lithotripsy can be made based on the determined calcification ablation state.

[0094] In one embodiment of the present invention, when performing calcification feature monitoring processing on a group of intravascular ultrasound images of a target blood vessel, the process includes:

[0095] Calcification region identification processing is performed sequentially on intravascular ultrasound images within the intravascular ultrasound image group. After calcification region identification processing is performed on all intravascular ultrasound images, the target calcification region, the target feature state of the target calcification region, and the calcification feature level of the target calcification region corresponding to the current intravascular ultrasound image group are determined.

[0096] The target characteristic state of the target calcification region includes at least the target calcification angle and / or the target calcification thickness;

[0097] The calcification characteristic grades include severe calcification, moderate calcification, and / or mild calcification.

[0098] In specific implementation, when performing calcification feature monitoring processing on intravascular ultrasound image groups, it specifically refers to sequentially performing calcification region identification processing on each frame of intravascular ultrasound image group. After performing calcification region identification processing on all intravascular ultrasound images in the intravascular ultrasound image group, the target calcification region, the target feature state of the target calcification region, and the calcification feature level of the target calcification region corresponding to the current intravascular ultrasound image group can be determined. The following corresponding explanations can be used as a reference for the calcification region identification processing on each frame of intravascular ultrasound image, as well as the determination of the target calcification region, the target feature state, and the calcification feature level.

[0099] The target characteristic state includes the target calcification angle and / or the target calcification thickness. Preferably, the target characteristic state may simultaneously include the target calcification angle and the target calcification thickness. The target calcification angle specifically refers to the calcification angle of the target calcified region, and the target calcification thickness specifically refers to the thickness of the target calcified region. The meanings of calcification angle and calcification thickness refer to angle and thickness in the general sense of this technical field. As can be seen from the above description, the calcification characteristic level characterizes the severity of the target calcified region. In one embodiment of the present invention, the calcification characteristic level includes severe calcification, moderate calcification, and / or mild calcification. The following is a specific explanation using the example of calcification characteristic levels including severe calcification, moderate calcification, and mild calcification.

[0100] In one embodiment of the present invention, configuring the impact crushing operating parameters of the IVL array element to match the calcification characteristic level of the determined calcified region includes:

[0101] When the calcification characteristic level of the target calcified area is severe calcification, configure the impact crushing working parameters of the IVL array element to put the IVL array element in high-energy mode.

[0102] When the calcification characteristic level of the target calcified area is moderate calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in medium energy mode.

[0103] When the calcification characteristic level of the target calcified area is light calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in low-energy mode.

[0104] The impact crushing operating parameters of the IVL array element group include at least the array element excitation voltage, excitation pulse width, excitation duty cycle, and energy density.

[0105] It should be understood that the calcification characteristic level is related to the target calcification angle and target calcification thickness of the target area. That is, different target calcification angles and thicknesses result in different calcification characteristic levels. In one embodiment, when the target calcification angle is greater than 180° and the target calcification thickness is greater than 1 mm, the corresponding calcification characteristic level is defined as severe calcification; when the target calcification angle is between 90° and 180° and the target calcification thickness is between 0.5 mm and 1 mm, the corresponding calcification characteristic level is defined as moderate calcification; and when the target calcification angle is less than 90° and the target calcification thickness is less than 0.5 mm, the corresponding calcification characteristic level is defined as mild calcification. Of course, the corresponding calcification characteristic level can also be determined based on the target calcification angle and target calcification thickness; examples will not be provided here.

[0106] When the target calcified region exhibits severe calcification, the IVL array elements should be in high-energy mode to rapidly perform shock wave lithotripsy on the target calcified region. In one embodiment, when the IVL array elements are configured in high-energy mode, the corresponding element excitation voltage can be 250-300V, the excitation duty cycle can be 5%-10%, the excitation pulse width can be 100μs, and the energy density can be 5-10J / cm². 2 Furthermore, when the IVL array elements are configured in medium-energy mode, in one embodiment, the corresponding array element excitation voltage can be 180-250V, the excitation duty cycle can be 1%-5%, the excitation pulse width can be 150μs, and the energy density can be 1-5J / cm². 2 When the IVL array elements are configured in low-energy mode, in one embodiment, the corresponding array element excitation voltage can be 100-180V, the excitation duty cycle can be 0.1%-1%, the excitation pulse width can be 200μs, and the energy density can be 0.1-1J / cm². 2 .

[0107] In the above description, the excitation pulse width refers to the time required for one impact using the IVL array elements. It is understood that a single intravascular shock wave lithotripsy (IVL) session may include multiple impacts; therefore, as described above, the time for a single IVL session can range from 10 ms to 100 ms.

[0108] Furthermore, after configuring the impact lithotripsy working parameters corresponding to the IVL array elements, in each working mode (high energy mode, medium energy mode, or low energy mode), in addition to matching with the calcification characteristic level, high-frequency modulated pulses can be used to trigger microbubble cavitation. After that, shock wave lithotripsy is performed to increase the effective effect of the shock wave on the target calcified area.

[0109] In one embodiment of the present invention, the stone impact termination condition includes an impact termination first condition and / or an impact termination second condition, wherein,

[0110] When the calcification ablation state meets either the first or second condition for impact termination, then the calcification characteristic change is matched with the impact termination condition of the crushed stone.

[0111] The first condition for terminating the impact includes an increase in vessel diameter of not less than the diameter increase threshold and a decrease in calcification thickness of not less than the thickness decrease threshold.

[0112] The second condition for terminating the impact includes the interruption of calcification continuity in the target calcified area, an increase in vessel diameter of not less than the diameter increase threshold, and a decrease in plaque echo intensity of not less than the echo intensity decrease threshold.

[0113] In order to further improve the accuracy and reliability of intravascular ultrasound lithotripsy, the lithotripsy impact termination conditions of the present invention may include a first impact termination condition and / or a second impact termination condition. Preferably, the lithotripsy impact termination conditions may include both the first impact termination condition and the second impact termination condition. The specific details of the first impact termination condition and the second impact termination condition are described below.

[0114] Specifically, when judging the calcification ablation status using the first condition for shock termination, the main focus is on the increase in vessel diameter and decrease in calcification thickness in the target calcified area after intravascular shock wave lithotripsy using the IVL array element. The threshold percentage for diameter increase can be 10%, and the threshold percentage for thickness decrease can be 20%. If, after one intravascular shock wave lithotripsy, the vessel diameter increases by ≥10% and the calcification thickness decreases by ≥20%, then the calcification ablation status is considered to meet the first condition for shock termination. For example, if two IVUS imaging sessions are performed using the IVUS array element to obtain two sets of intravascular ultrasound images, and intravascular shock wave lithotripsy is performed once using the IVL array element between the two IVUS imaging sessions, the two IVUS imaging sessions can be the first and second IVUS imaging sessions in the execution sequence. The intravascular shock wave lithotripsy performed in this case is the first intravascular shock wave lithotripsy performed after the first IVUS imaging. The initial state specifically refers to the state where the lithotripsy integration unit 100 is inserted, but IVUS imaging and intravascular shock wave lithotripsy have not been performed.

[0115] In the above example, after two IVUS imaging sessions, the vessel diameter of the target calcified region can be determined. Subsequently, the obtained vessel diameters can be compared to determine the percentage increase in vessel diameter after one intravascular shock wave lithotripsy session. Therefore: Where d represents the percentage increase in vessel diameter, R0 represents the initial vessel diameter, and R1 represents the vessel diameter after one intravascular shock wave lithotripsy (IVS) session. For calculations of the percentage increase in vessel diameter after subsequent IVS sessions, please refer to the instructions here; examples will not be provided here.

[0116] Furthermore, as explained above, after calcification feature monitoring processing, the target calcification thickness of the target calcification region can be obtained for each group of intravascular ultrasound images. Therefore, the calculation of the reduction in the modified thickness of the target calcification region can be referred to the above calculation instructions for the increase in vessel diameter, which will not be repeated here.

[0117] In practical implementation, commonly used techniques in this field can be employed to determine the vessel diameter of the target calcified region. For example, edge detection operators, such as the Canny operator, can be used for edge detection on each frame of intravascular ultrasound image. After edge detection, the inner boundary and adventitia of the vessel wall can be determined. The distance between the adventitia boundary and the inner boundary of the vessel wall is the corresponding basic diameter. Of course, other methods can also be used to determine the basic diameter of the target calcified region. As explained above, after each IVUS imaging, a group of intravascular ultrasound images containing multiple frames can be obtained. The arithmetic mean of all basic diameters belonging to the same IVUS imaging session is then calculated, and the arithmetic mean diameter value is used as the vessel diameter for comparison. Of course, other methods can also be used to determine the corresponding vessel diameter, which will not be illustrated here. In addition, the corresponding target calcification thickness can be calculated using the arithmetic mean method.

[0118] Specifically, when using the second condition for shock wave lithotripsy to determine the state of calcification ablation, the main focus is on assessing whether there is a disruption in the continuity of calcification, an increase in vessel diameter, and a decrease in plaque echo intensity after intravascular shock wave lithotripsy using IVL array elements. The percentage of increased vessel diameter can be referenced in the corresponding explanation of the first condition for shock wave lithotripsy mentioned above. The following provides a detailed explanation of the disruption in calcification continuity and the decrease in plaque echo intensity.

[0119] In practice, one feasible way to determine whether there is a break in the continuity of calcification is as follows:

[0120] Based on the same set of intravascular ultrasound images, it is determined whether the reduction ratio of the connected domain length of the target calcified region exceeds the threshold ratio of the reduction ratio of the connected domain length and whether the rate of change of the area of ​​the target calcified region exceeds the threshold ratio of the area change. Specifically, when the reduction ratio of the connected domain length of the target calcified region exceeds the threshold ratio of the reduction ratio of the connected domain length and the rate of change of the area of ​​the target calcified region exceeds the threshold ratio of the area change, it can be considered that there is a break in the continuity of calcification. Specifically, the threshold ratio of the reduction ratio of the connected domain length can be 30%, and the threshold ratio of the area change can be 25%.

[0121] Specifically, for each frame of intravascular ultrasound image, the basic connected region length of the basic calcified region is determined using techniques commonly used in this technology. Then, the arithmetic mean of the basic connected region lengths corresponding to all intravascular ultrasound images is taken to obtain the target connected region length of the target calcified region. After obtaining the target connected region length, the decrease in connected region length can be calculated using the method described above for calculating the increase in vessel diameter. It should be noted that the situation of the basic calcified region for each frame of intravascular ultrasound image can be referred to the corresponding explanation below. The basic connected region length of the basic calcified region is the connected region length in the usual sense; therefore, after determining the basic calcified region, the corresponding basic connected region length can be determined.

[0122] For the area change rate of the target calcified region, please refer to the calculation instructions for the reduction ratio of the connected region length mentioned above. For each frame of intravascular ultrasound image, the area of ​​the corresponding basic calcified region is determined by the technical means commonly used in this field. Then, based on the area of ​​the basic calcified region, the area of ​​the corresponding target calcified region can be calculated by the arithmetic mean. After obtaining the area of ​​the target calcified region, the area change rate of the target calcified region can be calculated.

[0123] For plaque echo intensity, it specifically refers to the average grayscale value of all pixels within the target calcified region. That is, for each frame of intravascular ultrasound image, a corresponding basic plaque echo intensity can be calculated. Then, the arithmetic mean of the basic plaque echo intensities corresponding to the same intravascular ultrasound image group is taken as the target plaque echo intensity corresponding to that group. After calculating the target plaque echo intensity, it can be compared with the initial state of the plaque echo intensity of the target calcified region to determine the percentage reduction in plaque echo intensity. In practice, the echo intensity reduction threshold can be 15%, but other values ​​can also be used, depending on the specific requirements.

[0124] In one embodiment of the present invention, the process of identifying calcified regions in a group of intravascular ultrasound images includes:

[0125] A calcification segmentation model was constructed and deployed within the gravel control unit 200.

[0126] When performing calcification region identification processing on intravascular ultrasound images, a calcification segmentation model is used to segment and identify each pixel in the intravascular ultrasound image. After segmentation and identification, the mask type of each pixel is determined, and the intravascular ultrasound image is segmented into basic calcification region and background region based on the mask type of each pixel. The mask type of each pixel is either foreground mask or background mask.

[0127] Based on the basic calcification regions of all intravascular ultrasound images, a target calcification region corresponding to the current intravascular ultrasound image is generated.

[0128] For each basic calcification region in an intravascular ultrasound image, calculate the basic calcification angle and basic calcification thickness corresponding to the basic calcification region.

[0129] Based on all the basic calcification angles, generate the target calcification angle;

[0130] Based on all the basic calcification thicknesses, the target calcification thickness is generated.

[0131] To improve the accuracy and reliability of calcification region identification, a calcification segmentation model should be constructed. After constructing the required calcification segmentation model, it should be deployed within the lithotripsy control unit 200. Subsequently, the calcification segmentation model can be used to perform calcification region identification processing on each frame of intravascular ultrasound image. During calcification region identification processing, each pixel of the intravascular ultrasound image is segmented and identified. After segmentation and identification, the mask type of each pixel is determined. The mask type can be a foreground mask or a background mask. Thus, each pixel can be segmented and identified as a foreground mask or a background mask. Subsequently, pixels based on the foreground mask can form basic calcification regions, and pixels based on the background mask can form background regions. This segmentation and identification method can achieve binary segmentation of intravascular ultrasound images.

[0132] Understandably, after identifying calcified regions in each frame of intravascular ultrasound (IVUS) image, a corresponding basic calcified region can be obtained. In practice, for each group of IVUS images, one basic calcified region can be selected as the target calcified region, and subsequent IVUS imaging will be based on this selected region. Furthermore, for IVUS images belonging to the same group, the corresponding target calcified angle can be calculated based on all basic calcified angles, and the target calcified thickness can be generated based on all basic calcified thicknesses. Specifically, the basic calcified angle is the calcified angle corresponding to the basic calcified region in each frame of IVUS image, and similarly, the basic calcified thickness is the calcified thickness corresponding to the basic calcified region in each frame of IVUS image.

[0133] In practice, when generating the target calcification thickness based on all basic calcification thicknesses, the arithmetic mean of all target calcification thicknesses can be calculated, and the result of the arithmetic mean calculation can be used as the target calcification thickness. Furthermore, the target calcification angle can be generated in the same way.

[0134] As explained above, the lithotripsy integration unit 100 remains fixed after vascular intervention. That is, when configuring the IVUS array for IVUS imaging, it images the same region of the target vessel. Therefore, the basic calcification region generated based on each frame of intravascular ultrasound image can characterize the same region within the target vessel. Thus, when any basic calcification region is selected as the target calcification region, it can still accurately characterize the target calcification region corresponding to the set of intravascular ultrasound images. The target calcification angle and thickness calculated using the above method can improve the accuracy of the calcification angle and thickness calculations, thereby improving the accuracy and reliability of generating calcification feature levels.

[0135] Generally, after the segmentation and identification described above, each frame of intravascular ultrasound image will contain a basic calcified region. Understandably, due to factors such as the contact state of the IVUS array elements, artifacts may exist within this basic calcified region. To correct these artifacts, artifact detection and correction processing should be performed on the obtained basic calcified region. Specifically,

[0136] When performing artifact detection, the average gray value of the basic calcified region and the edge gradient of the basic calcified region are calculated. If the calculated average gray value of the region is less than 20% of the global average gray value and the edge gradient is greater than 50, then the basic calcified region is considered to have artifacts. Otherwise, the basic calcified region is considered not to have artifacts.

[0137] Specifically, the regional average grayscale value is the average of the grayscale values ​​of all pixels within the basic calcified region, and the global average grayscale value is the average of the grayscale values ​​of all pixels in the corresponding intravascular ultrasound image. The corresponding intravascular ultrasound image specifically refers to the intravascular ultrasound image on which the segmentation and identification of the basic calcified region depend. When artifacts are detected, commonly used techniques in this technical field can be employed to correct the artifacts in the basic calcified region. The specific methods for artifact correction are consistent with existing technologies and will not be illustrated here.

[0138] In one embodiment of the present invention, the calcification segmentation model includes a basic segmentation network and a group of acoustic and shadow units adapted and connected to the basic segmentation network, wherein,

[0139] The basic segmentation network includes an encoder network and a decoder network adapted and connected to the encoder network. The encoder network includes several encoding layers, and the decoder network includes several decoding layers. The number of decoding layers in the decoder network is the same as the number of encoding layers in the encoder network, and the encoder network and decoder network are U-shaped after being adapted and connected.

[0140] The sound and shadow unit group includes several sound and shadow units. The number of sound and shadow units is less than the number of coding layers in the encoder network. Each sound and shadow unit acquires the mid-to-deep features generated by a coding layer in the encoder network and performs calcification-sound and shadow association on the acquired mid-to-deep features to generate a sound and shadow target feature map. The sound and shadow target feature map is used to model the directional dependency relationship between the calcified strong echo region and the sound and shadow region behind the calcified strong echo region.

[0141] The sound and shadow target feature map generated by each sound and shadow unit is loaded into the decoding layer that corresponds to the current coding layer to supplement the semantic understanding of the strong echo-low echo association of the decoding layer.

[0142] To perform the aforementioned calcification region identification processing on intravascular ultrasound images, the calcification segmentation model can include a basic segmentation network and a set of acoustic shadowing units. The basic segmentation network can be a commonly used U-Net network. When using a U-Net network, the basic segmentation network can include an encoder network and a decoder network. The encoder network includes several encoding layers, and the decoder network includes several decoding layers. The decoder network and encoder network are adapted and connected to form the structure of a U-Net network. Figure 6 The figure shows an embodiment of a basic segmentation network, in which the encoder network includes four encoding layers, and the decoder network also includes four decoding layers.

[0143] In practical implementation, the coding and decoding layers can adopt existing common forms. For two adjacent coding layers, the output of the previous coding layer is downsampled and used as the input of the next coding layer. At the same time, the output of each coding layer is also connected to the corresponding decoding layer via skip connections, such as... Figure 5 In the embodiment shown, the encoding layer of the first layer and the decoding layer of the first layer are skipped connections, and the encoding layer of the first layer is the encoding layer for receiving loaded intravascular ultrasound images.

[0144] Each encoding layer performs the same encoding process on the input feature map. Figure 6 In the diagram, US0 is the intravascular ultrasound image to be processed for calcification region identification. Within the first coding layer, the intravascular ultrasound image undergoes two coding convolution processes. The first coding convolution generates feature map E0, and the second coding convolution generates feature map E1, which is the coded feature map. Generally, the feature dimension of an intravascular ultrasound image can be 512*512*1, the feature dimension of feature map E0 is 512*512*64, and the feature dimension of feature map E1 is 512*512*64. As explained above, feature map E1 is connected to the first decoding layer, and it is downsampled to generate feature map E2, which has a feature size of 256*256*64. Figure 6 In the diagram, MP1 is the downsampling unit that performs downsampling processing on the feature map E1. Furthermore, Figure 6 In the diagram, MP2, MP3, and MP4 are the corresponding downsampling units.

[0145] Depend on Figure 6 As explained above, along the direction of downsampling, the coding layer before downsampling is the coding layer of the previous layer, and the coding layer after downsampling is the coding layer of the next layer. For details on the coding convolution and downsampling processes performed on other coding layers, please refer to [reference needed]. Figure 6 The corresponding explanations mentioned above will not be repeated here.

[0146] To avoid misjudgments caused by relying solely on strong echo (calcified region) signals, in one embodiment of the present invention, a sound shadow unit group is further provided within the calcification segmentation model. The sound shadow unit group may include several sound shadow units. Generally, the number of sound shadow units is less than the number of coding layers in the encoder network, and also less than the number of decoding layers in the decoder network. Specifically, one sound shadow unit can capture mid-to-deep features generated by one coding layer within the encoder network. Therefore, the coding layer corresponding to the mid-to-deep features should be located in the mid-to-deep layer, such as... Figure 6 In the embodiments shown, the third and fourth coding layers can be regarded as coding layers in the middle and deep layers. When the encoder network is in other cases, the corresponding coding layers in the middle and deep layers can be obtained. Examples will not be given here.

[0147] Figure 6 The figure shows an embodiment where the audio-visual unit group includes only one audio-visual unit. In the figure, the audio-visual unit acquires mid-to-deep features generated by the fourth coding layer. When the audio-visual unit group includes multiple audio-visual units, refer to [the following text is missing from the original] Figure 6 And the explanation here, such as Figure 6 In the illustrated embodiment, when two audio-visual units exist, the other audio-visual unit should acquire the mid-to-deep features generated by the third coding layer. Figure 6 As explained above, the mid-to-deep features generated by the coding layer specifically refer to the feature maps generated after downsampling the output of the mid-to-deep coding layer, such as... Figure 6 In the embodiment shown, the fourth coding layer generates a coding feature map E10, which is then downsampled by the downsampling unit MP4 to generate a feature map E11. At this time, the feature map E11 is used as a mid-to-deep feature. Other cases can be referred to the description here.

[0148] In one embodiment of the present invention, each acoustic shadowing unit performs calcification-acoustic shadowing correlation on the acquired mid-to-deep features to generate an acoustic shadowing target feature map, thereby using the acoustic shadowing target feature map to model the directional dependency between the calcified strong echo region and the acoustic shadowing region behind the calcified strong echo region. Figure 6 In the embodiment shown, E14 is the generated acoustic shadow target feature map. The calcified strong echo region is the calcified region; the acoustic shadow region behind the calcified strong echo region is generally the low echo region. The acoustic shadow specifically refers to the characteristics of the low echo region.

[0149] For each sound-shadow target feature map generated by the sound-shadow unit, the sound-shadow target feature map should be loaded into the decoding layer corresponding to the current coding layer to supplement the decoding layer's semantic understanding of the strong echo-low echo association. Specifically, the corresponding coding layer refers to the variable coding layer directly corresponding to the mid-to-deep features obtained by the sound-shadow unit. It can be understood that after performing calcification-sound-shadow association on the obtained mid-to-deep features through the sound-shadow unit, the sound-shadow behind the calcified region can be used for auxiliary segmentation, optimizing segmentation accuracy, thereby improving the accuracy and reliability of segmentation recognition.

[0150] In one embodiment of the present invention, each coding layer in the encoder network outputs a coding feature map, wherein the mid-to-deep features are generated by downsampling the coding feature map output by the coding layer located in the mid-to-deep position;

[0151] When the acquired mid-to-deep features are generated by downsampling the encoded feature map of the deepest encoding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is upsampled, and in the corresponding decoding layer, the sound and shadow upsampled feature map and the corresponding encoded feature map are concatenated.

[0152] When the acquired mid-to-deep features are generated by downsampling the coding feature map of the mid-domain coding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is added to the corresponding decoding feature map, and the sound and shadow decoding feature map is generated after the addition.

[0153] The audio-visual decoding feature map is upsampled to generate an upsampled audio-visual decoding feature map. The generated upsampled audio-visual decoding feature map is then concatenated with the corresponding encoded feature map output by the encoding layer. The decoded feature map is the feature map generated by the previous decoding layer.

[0154] Specifically, the method for performing downsampling can be found in [reference needed]. Figure 6 And the corresponding explanations above. Furthermore, as can be seen from the above explanations, a sound and shadow unit group may include one or more sound and shadow units, and a sound and shadow unit may be connected to a coding layer located in a mid-to-deep layer. The following will combine... Figure 6 Examples are given to illustrate the connections between the audio-visual unit and the corresponding encoding and decoding layers.

[0155] Figure 6 In the embodiment shown, the sound-shadow unit corresponds to the fourth coding layer, which is the deepest coding layer. After the fourth coding layer generates a coded feature map, it is downsampled by the downsampling unit MP4 to generate feature map E11. The sound-shadow unit performs calcification-sound-shadow association on feature map E11 to generate a sound-shadow target feature map. Figure 6In the diagram, E14 represents the generated sound and shadow target feature map. Subsequently, the sound and shadow target feature map E14 is upsampled and then loaded into the fourth decoding layer. Figure 6 In the diagram, UP1 to UP4 are the corresponding upsampling processing units, which perform upsampling processing. Specifically, in the fourth decoding layer, the encoded feature map E10 output from the fourth encoding layer is concatenated with the audio-visual target feature map E14 to form feature map D1. Then, in the fourth decoding layer, feature map D1 undergoes decoding convolution to generate feature map D2. Feature map D2 then undergoes decoding convolution to generate feature map D3, which serves as the decoded feature map for the fourth decoding layer. During decoding, upsampling is performed using upsampling unit UP2. Then, in the third decoding layer, the upsampled feature map is concatenated with the encoded feature map E7 to form feature map D4. For details on decoding processes at different layers within the decoder network, please refer to [reference needed]. Figure 6 And this is an explanation, which will not be listed here again.

[0156] As explained above, the audio-visual unit can also correspond to a mid-to-deep coding layer. Here, the mid-domain coding layer refers to any mid-to-deep coding layer other than the deepest coding layer, such as... Figure 6 In the illustrated embodiment, this can correspond to the coding layer of the third layer. Specifically, the coded feature map output by the middle-domain coding layer is downsampled to generate a feature map loaded into the current audio-visual unit, and the audio-visual unit can generate the corresponding audio-visual target feature map. In order to enter the corresponding decoding layer, the audio-visual target feature map should be added to the decoded upsampled feature map, and then upsampled after addition. Subsequently, in the decoding layer, it is concatenated with the coded feature map of the corresponding coding layer. For example, if the middle-domain coding layer is... Figure 6 When the third encoding layer is used, the corresponding decoding layer is the third decoding layer.

[0157] Using the middle-domain coding layer as Figure 6 Taking the third coding layer as an example, the third coding layer outputs feature map E7. After downsampling feature map E7, feature map E8 is obtained. At this time, feature map E8 is loaded into the sound shadow unit corresponding to the third coding layer. The sound shadow unit can generate sound shadow target feature map E16. It should be understood that the form of connection and cooperation between sound shadow target feature map E16, sound shadow unit and the third coding layer is not in the... Figure 6 As shown in the image.

[0158] The generated sound and shadow target feature map E16 is added to the decoded feature map. Figure 6In the process, the decoded feature map is feature map D3. The sound and shadow target feature map E16 is added to feature map D3 to generate the sound and shadow decoding feature map. Subsequently, the sound and shadow decoding feature map undergoes upsampling processing, such as... Figure 6 The upsampling unit UP2 performs upsampling processing, generating an upsampled feature map for audio-visual decoding. Then, the upsampled feature map is concatenated with feature map E7, generating feature map D4. This completes the loading of mid-to-deep features, processed by the audio-visual unit, onto the corresponding decoding layer.

[0159] Depend on Figure 6 As explained above, the feature dimension of feature map E8 is 64*64*256. Referring to the generation instructions for feature map E14, the feature dimension of feature map E16 should be 64*64*512. The feature dimension of feature map D3 is 64*64*512. Therefore, feature map E16 can be added to feature map D3 to obtain the audio-visual decoding feature map, which has a feature dimension of 64*64*512. Then, the audio-visual decoding feature map is upsampled to generate the upsampled audio-visual decoding feature map, which has a feature dimension of 128*128*256. Within the third decoding layer, the upsampled audio-visual decoding feature map and feature map E7 are concatenated, and after channel bonding, feature map D4 with a feature dimension of 128*128*512 can be generated.

[0160] Figure 6 The diagram illustrates one embodiment of the decoding layer within the decoder network. As shown, the decoding layer performs two decoding convolution processes. The first decoding convolution process reduces the number of channels by half while maintaining the same feature map size. The second decoding convolution process maintains the same feature map size as the one after the first decoding convolution process. After decoding by the first decoding layer, an output convolution process is performed. This output convolution process ensures that the size of the output feature map matches the feature dimensions of the intravascular ultrasound image.

[0161] In one embodiment of the present invention, the sound and shadow unit includes a sound and shadow module and a sound and shadow post-processing module, wherein,

[0162] The audio-visual module includes a semantic alignment enhancement module and a directional dependency modeling unit connected in sequence, wherein,

[0163] The semantic alignment enhancement module receives the loaded mid-to-deep features and performs semantic alignment enhancement processing on the mid-to-deep features. After semantic alignment enhancement processing, the contrast between the calcified strong echo region and the sound shadow region is focused, and a mid-to-deep semantic alignment enhancement feature map is generated.

[0164] The spatial orientation capture process of the mid-to-deep semantic alignment enhancement feature map is performed by the directional dependency modeling unit to capture the directional propagation features of the calcified strong echo region and the sound shadow region, and to generate the sound shadow basic feature map. The feature dimension of the sound shadow basic feature map is consistent with the feature dimension of the mid-to-deep features.

[0165] The audio-visual post-processing module performs at least feature processing on the basic audio-visual feature map to generate the audio-visual target feature map after feature processing.

[0166] To enable calcification-sound-shadow correlation of the acquired mid-to-deep features, the sound-shadow unit may include a sound-shadow module and a sound-shadow post-processing module. Figure 6 The figure shows an embodiment of the sound and shadow unit. In the figure, feature map E12 is the basic sound and shadow feature map output by the sound and shadow module. The feature dimension of the basic sound and shadow feature map is consistent with the feature dimension of the mid-deep feature E11. Figure 6 The figure also illustrates one embodiment of the audio-visual post-processing module. The module may include two post-convolutional units. Specifically, the two post-convolutional units perform convolution operations on the basic audio-visual feature map. After the first convolution operation, feature map E13 is generated. Feature map E13 is then further convolved by the post-convolutional units to generate feature map E14. Figure 6 In the image, the feature dimension of feature map E13 is 32*32*1024, and the feature dimension of feature map E14 is consistent with that of feature map E13. Therefore, it can be seen that the feature processing performed on the basic feature map of sound and shadow by the sound and shadow post-processing module can be a convolution operation.

[0167] In specific implementation, the sound and shadow module may include a semantic alignment enhancement module and a directional dependency modeling unit. Specifically, for the acquired mid-to-deep features, the sound and shadow module first uses the semantic alignment enhancement module to perform semantic alignment enhancement processing, and after semantic alignment enhancement processing, it can generate a mid-to-deep semantic alignment enhancement feature map. Figure 7 The figure illustrates one embodiment of the semantic alignment enhancement module. As shown in the figure, the semantic alignment enhancement module may include a 1×1 convolution module and an embedded lightweight semantic attention module. The 1×1 convolution module compresses the number of channels of the mid-to-deep features to align the semantic feature dimensions of the calcified region and the sound shadow region. Subsequently, the embedded lightweight semantic attention module enhances the semantic information of the feature map output by the 1×1 convolution module to highlight the channels of strong echo-low echo association. The highlighted channels may be the edges of the calcified region and the sound shadow gradient region, thereby focusing the contrast between the calcified strong echo region and the sound shadow region. The embedded lightweight semantic attention module can output a mid-to-deep semantic alignment enhancement feature map.

[0168] Figure 7 In this context, conv 1×1 is a 1×1 convolutional module. Figure 7 The document also illustrates an embodiment of embedding a lightweight semantic attention module using a selayer. For details on embedding a lightweight semantic attention module using a selayer, please refer to the selayer documentation; it will not be repeated here.

[0169] The mid-to-deep semantic alignment enhancement feature map output by the semantic alignment enhancement module can be spatially captured by the directional dependency modeling unit to capture the directional propagation features of the calcified strong echo region and the sound shadow region, and generate the basic sound shadow feature map. Figure 7 The diagram also illustrates one embodiment of a directional dependency modeling unit. As shown in the figure, the directional dependency modeling unit may include a difference mapping module, a direction-aware module, and a radial spatial attention module connected in sequence. The following section will discuss... Figure 7 The case of the directional dependency modeling unit shown in the figure will be explained in detail.

[0170] Specifically, the difference mapping module performs feature selection and difference mapping on the mid-to-deep semantic alignment enhancement feature map, and generates a strong echo-low echo difference feature map. Figure 7 The image shows an embodiment of feature selection and difference mapping, specifically:

[0171] D=σ(Conv 1×1 (F in ))Θtanh(Conv 1×1 (F in ))

[0172] Where D is the generated strong echo-low echo difference feature map, σ is the Sigmoid function, and Conv 1×1 For 1×1 convolution, Θ is the pointwise multiplication operation, and F is the value of F. in This is a feature map for enhancing semantic alignment in the mid-to-deep layers, and tanh is the hyperbolic tangent operation.

[0173] Depend on Figure 7 As explained above, feature selection and difference mapping involve two branches. In one branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a sigmoid function operation. In the other branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a tanh operation. Afterward, the results from both branches are multiplied point-by-point. Figure 7 CF1 in the diagram is a point-by-point multiplier, which can output a strong echo-low echo difference feature map.

[0174] After generating the strong echo-low echo difference feature map, the direction perception module performs direction perception processing on the strong echo-low echo difference feature map so that a direction perception feature map can be generated after the direction perception processing. Figure 7The figure illustrates one embodiment of the orientation sensing module. In the figure, the strong echo-low echo difference feature map is processed by directional convolution using a 3×3 convolution with angle weights, resulting in:

[0175]

[0176] Among them, F dir This is a direction-aware feature map generated by directional convolution. This is a directional convolution process for a 3×3 convolution with angle weights.

[0177] Figure 7 In the diagram, `conv 3x3` represents a 3×3 convolution, and the 3×3 matrix above it is the weight matrix for different angles of the convolution kernel. Specifically, the weight values ​​of the weight matrix should be set according to the polar coordinate characteristics when IVUS elements are used to form an image (calcification is distributed along the axial direction of the blood vessel, and acoustic shadowing is distributed along the radial direction). Figure 7 In setting the weight values, the weight value can be 0.6 in the axial direction (0°, 180°) to enhance the capture of the continuity of calcified areas; and the weight value can be 1.2 in the radial direction (90°, 270°) to highlight the directional propagation of sound shadows along the radial direction.

[0178] After generating the direction-aware feature map, the radial spatial attention module first performs radial spatial attention processing to generate a radial spatial attention feature map. Then, the radial spatial attention feature map is multiplied point-by-point with the direction-aware feature map to generate a basic sound-image feature map. Specifically, the radial spatial attention feature map includes several radial spatial attention features, which are:

[0179]

[0180] Among them, Att radial (x,y) represents the radial spatial attention feature of the pixel at coordinate (x,y), (x,y) represents the coordinate position of the pixel on the orientation-aware feature map, r(x,y) represents the radial distance from pixel (x,y) to the center of the blood vessel, and α represents the distance weighting coefficient.

[0181] In the radial spatial attention processing described above, the weight distribution is calculated based on the radial distance from the pixel to the center of the blood vessel. This allows the calcification segmentation model to focus more on feature associations along the radial direction of the blood vessel (extending outward from the center). The radial distance is the radius dimension in polar coordinates. It is understood that when calculating the radial distance, the position coordinates of the blood vessel center should first be determined using techniques commonly used in this technical field. The distance weight coefficient α is used to control the intensity of the radial distance's influence on attention. The larger the distance weight coefficient α, the higher the attention weight of distant pixels. The distance weight coefficient α can generally be set to 0.8.

[0182] For the radial spatial attention feature expression calculated above, the numerator represents the radial distance of a single pixel mapped to an exponential weight, highlighting the contribution of pixels at greater distances (further outwards along the radial direction). The denominator normalizes the exponential weights of all pixels, ensuring that the sum of the attention weights is 1, conforming to the probability distribution characteristics. Based on the calculated radial spatial attention features, a corresponding radial spatial attention feature map can be generated. The method for generating the radial spatial attention feature map can be consistent with existing methods and will not be elaborated here.

[0183] When generating the basic sound and shadow feature map, we have: Fdep = F dir ΘAtt radial Where Fdep is the basic feature map of sound and shadow, and Att radial For radial spatial attention features, Θ represents pointwise multiplication. As can be seen from the process of generating the basic sound-shadow feature map in this invention, the generated basic sound-shadow feature map can model the directional dependency between the calcified strong echo region and the low echo region behind it, highlighting the feature weight of the low echo region behind the calcified region.

[0184] Figure 7 In this context, Fenc represents mid-to-deep features, the semantic alignment module is the semantic alignment enhancement module, the difference mapping module is the difference mapping module, the orientation-aware convolution module is the orientation-aware module, and the radial attention module is the radial spatial attention module. CF2 is a pointwise multiplier, which outputs a basic sound and shadow feature map.

[0185] The calcification segmentation model described above can be constructed and generated in the following manner. Specifically, a feasible construction method includes:

[0186] Construct a basic model for calcification segmentation and build a basic model training dataset for training the basic model for calcification segmentation.

[0187] Configure the training conditions for training the basic model of the calcification segmentation model, so as to train the basic model of calcification segmentation model using the basic model training dataset under the configured training conditions;

[0188] Once the training of the basic calcification segmentation model reaches the target state, a calcification segmentation model is generated based on the basic calcification segmentation model.

[0189] It should be understood that the basic calcification segmentation model can be consistent with the calcification segmentation model described above. Therefore, the basic segmentation model can be constructed with reference to the description of the calcification segmentation model described above. The basic model training dataset includes several training samples. Each training sample includes an intravascular training image and a pixel-level segmentation mask labeled on the intravascular training image. The intravascular training image can be obtained by collecting clinical ultrasound images of coronary artery scenes from different hospitals. In specific implementation, the intravascular training images in the basic model training dataset should include different degrees of calcification and vessel diameters.

[0190] To enhance the training samples, geometric transformations can be applied to the collected clinical ultrasound images. These transformations can include rotation, scaling, and / or flipping. The methods and processes for geometric transformation of clinical ultrasound images can be selected as needed and will not be elaborated here. When performing pixel-level segmentation mask annotation on clinical ultrasound images, 2-3 interventional cardiovascular experts can independently annotate the images, resolving disagreements through a consensus mechanism. The specific methods and processes for pixel-level segmentation mask annotation will not be elaborated here. It should be noted that pixel-level segmentation mask annotation specifically refers to foreground mask annotation of pixels belonging to calcified regions on clinical ultrasound images.

[0191] The configured training conditions may include an optimizer and training parameters. Specifically, the optimizer can use either the Adam or Ranger optimizer. Training parameters may include: an initial learning rate set to 1e-4 to 1e-3, dynamically adjusted using cosine annealing or the ReduceLROnPlateau strategy; furthermore, considering the image resolution of intravascular ultrasound images in IVUS imaging (typically 512×512 or higher), the batch size is set to 4 to 8 to avoid GPU memory overflow. Additionally, training conditions may include a training loss function and / or other necessary parameters. The training loss function is explained in detail below.

[0192] Lz=λ1*L Dice +λ2*L fz

[0193] Where Lz is the training loss function, λ1 and λ2 are weight parameters, and L Dice For Dice's loss, L fz To mitigate losses.

[0194] Specifically, the values ​​of weight parameters λ1 and λ2 can be selected as needed, and the auxiliary loss L... fz The binary cross-entropy loss L can be used. BCE and / or combined with focal loss L Focal Loss L through Dice Dice It can measure the segmentation accuracy of calcified regions. Furthermore, when the auxiliary loss L... fzIncluding binary cross-entropy loss L BCE When the auxiliary loss L can be used to measure the classification probability, fz Including combined coke loss L Focal At that time, it can effectively solve the problem of category imbalance between calcified areas and background. The following section discusses the Dice loss L... Dice Binary cross-entropy loss L BCE Combined with coke loss L Focal The specific details will be explained below.

[0195] For Dice loss L Dice Then we have:

[0196]

[0197] Where N is the number of training samples in a batch, H×W is the resolution of the intravascular training image, and p ij Let g be the probability that the j-th pixel in the i-th intravascular training image is a foreground mask. ij Let be the ground truth mask label for the j-th pixel of the i-th intravascular training image. To prevent smooth terms with a denominator of 0, It can generally be 1e -5 .

[0198] For the binary cross-entropy loss L BCE Then we have:

[0199]

[0200] Here, log is the base-10 logarithmic operation.

[0201] For the combined coke loss L Focal Then we have:

[0202]

[0203] Where β and δ are weighting coefficients.

[0204] Specifically, the weighting coefficient β can be 0.75, which can be used to re-minimize the foreground mask that is difficult to classify, and the weighting coefficient δ can be 2, which is used to reduce the weight of the background mask that is easy to classify.

[0205] It should be noted that the configured training conditions should generally also include the number of training rounds. When the training of the basic calcification segmentation model reaches the required number of training rounds, or when the value of the training loss function no longer decreases, the training of the basic calcification segmentation model can be considered to have reached the target state. The training termination condition can be selected as needed. When the training termination condition is reached, the target basic calcification segmentation model can be determined according to the setting of the training termination condition. Based on the determined basic calcification segmentation model, it can be configured as the calcification segmentation model. At this time, the calcification segmentation model is constructed. The method of deploying the calcification segmentation model in the gravel control unit 200 can be consistent with the existing technology, and will not be elaborated here.

Claims

1. An intravascular ultrasound lithotripsy system integrating IVUS and IVL, characterized in that, The intravascular ultrasonic lithotripsy system includes: The lithotripsy integrated unit includes at least an imaging tube assembly and an ultrasonic array element component adapted and connected to the imaging tube assembly, wherein... The ultrasound array assembly includes at least an IVUS array for acquiring intravascular ultrasound images and an IVL array for performing intravascular shock wave lithotripsy. The IVL array and the IVUS array are integrated on the same backing and mounted on the imaging tube assembly through the backing. The lithotripsy control unit performs calcification feature monitoring on the intravascular ultrasound images of the target vessel acquired via IVUS array elements. This calcification feature monitoring process determines at least the calcification characteristic level and corresponding calcification ablation status of the target calcified region. When the calcification ablation state matches the lithotripsy termination condition, the lithotripsy control unit stops the IVL array group from performing intravascular shock wave lithotripsy; otherwise, the IVL array group is configured to enter intravascular shock wave lithotripsy operation. When configuring the IVL array element to perform intravascular shock wave lithotripsy, the lithotripsy control unit shall at least configure the shock wave lithotripsy parameters of the IVL array element to match the calcification characteristic level of the target calcified area.

2. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 1, characterized in that: The imaging tube assembly includes an inner tube and an integrated outer tube fitted onto the inner tube, wherein... The backing sleeve is placed on the inner tube, and the backing sleeve, IVUS array elements, and IVL array elements are covered by a balloon assembled on the inner tube. The IVUS elements in the IVUS array group and the IVL elements in the IVL array group are arranged in alternating rings on the backing. The IVUS elements and IVL elements are coplanar on the backing, and the ultrasonic emission direction of the IVL elements is radially aligned with the detection direction of the IVUS elements. The integrated outer tube is at least fitted over the tail end of the inner tube, and the balloon is located between the head end of the integrated outer tube and the head end of the inner tube.

3. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 1, characterized in that, When performing calcification feature monitoring on intravascular ultrasound images of the target blood vessel, the following steps are included: Calcification region identification processing is performed sequentially on intravascular ultrasound images within the intravascular ultrasound image group. After calcification region identification processing is performed on all intravascular ultrasound images, the target calcification region, the target feature state of the target calcification region, and the calcification feature level of the target calcification region corresponding to the current intravascular ultrasound image group are determined. The target characteristic state of the target calcification region includes at least the target calcification angle and / or the target calcification thickness; The calcification characteristic grades include severe calcification, moderate calcification, and / or mild calcification.

4. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 3, characterized in that, When configuring the impact crushing parameters of the IVL array to match the calcification characteristic level of the identified calcified region, the following are included: When the calcification characteristic level of the target calcified area is severe calcification, configure the impact crushing working parameters of the IVL array element to put the IVL array element in high-energy mode. When the calcification characteristic level of the target calcified area is moderate calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in medium energy mode. When the calcification characteristic level of the target calcified area is light calcification, configure the impact crushing working parameters of the IVL array element to make the IVL array element in low-energy mode. The impact crushing operating parameters of the IVL array element group include at least the array element excitation voltage, excitation pulse width, excitation duty cycle, and energy density.

5. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 3, characterized in that, The conditions for terminating the impact of the crushed stone include a first condition for termination and / or a second condition for termination, wherein... When the calcification ablation state meets either the first or second condition for impact termination, then the calcification characteristic change is matched with the impact termination condition of the crushed stone. The first condition for terminating the impact includes an increase in vessel diameter of not less than the diameter increase threshold and a decrease in calcification thickness of not less than the thickness decrease threshold. The second condition for terminating the impact includes the interruption of calcification continuity in the target calcified area, an increase in vessel diameter of not less than the diameter increase threshold, and a decrease in plaque echo intensity of not less than the echo intensity decrease threshold.

6. The integrated IVUS and IVL intravascular ultrasound lithotripsy system according to any one of claims 3 to 5, characterized in that, When performing calcification region identification processing on intravascular ultrasound images, the following steps are included: A calcification segmentation model was constructed and deployed within the gravel control unit. When performing calcification region identification processing on intravascular ultrasound images, a calcification segmentation model is used to segment and identify each pixel in the intravascular ultrasound image. After segmentation and identification, the mask type of each pixel is determined, and the intravascular ultrasound image is segmented into basic calcification region and background region based on the mask type of each pixel. The mask type of each pixel is either foreground mask or background mask. Based on the basic calcification regions of all intravascular ultrasound images, a target calcification region corresponding to the current intravascular ultrasound image is generated. For each basic calcification region in an intravascular ultrasound image, calculate the basic calcification angle and basic calcification thickness corresponding to the basic calcification region. Based on all the basic calcification angles, generate the target calcification angle; Based on all the basic calcification thicknesses, the target calcification thickness is generated.

7. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 6, characterized in that, The calcification segmentation model includes a basic segmentation network and a group of acoustic-shadow units adapted and connected to the basic segmentation network, wherein... The basic segmentation network includes an encoder network and a decoder network adapted and connected to the encoder network. The encoder network includes several encoding layers, and the decoder network includes several decoding layers. The number of decoding layers in the decoder network is the same as the number of encoding layers in the encoder network, and the encoder network and decoder network are U-shaped after being adapted and connected. The sound and shadow unit group includes several sound and shadow units. The number of sound and shadow units is less than the number of coding layers in the encoder network. Each sound and shadow unit acquires the mid-to-deep features generated by a coding layer in the encoder network and performs calcification-sound and shadow association on the acquired mid-to-deep features to generate a sound and shadow target feature map. The sound and shadow target feature map is used to model the directional dependency relationship between the calcified strong echo region and the sound and shadow region behind the calcified strong echo region. The sound and shadow target feature map generated by each sound and shadow unit is loaded into the decoding layer that corresponds to the current coding layer to supplement the semantic understanding of the strong echo-low echo association of the decoding layer.

8. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 7, characterized in that, Each coding layer in the encoder network outputs a coding feature map, wherein the mid-to-deep features are generated by downsampling the coding feature map output by the coding layer located in the mid-to-deep position; When the acquired mid-to-deep features are generated by downsampling the encoded feature map of the deepest encoding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is upsampled, and in the corresponding decoding layer, the sound and shadow upsampled feature map and the corresponding encoded feature map are concatenated. When the acquired mid-to-deep features are generated by downsampling the coding feature map of the mid-domain coding layer, after generating the sound and shadow target feature map, the sound and shadow target feature map is added to the corresponding decoding feature map, and the sound and shadow decoding feature map is generated after the addition. The audio-visual decoding feature map is upsampled to generate an upsampled audio-visual decoding feature map. The generated upsampled audio-visual decoding feature map is then concatenated with the corresponding encoded feature map output by the encoding layer. The decoded feature map is the feature map generated by the previous decoding layer.

9. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 7, characterized in that, The audio-visual unit includes an audio-visual module and an audio-visual post-processing module, wherein... The audio-visual module includes a semantic alignment enhancement module and a directional dependency modeling unit connected in sequence, wherein, The semantic alignment enhancement module receives the loaded mid-to-deep features and performs semantic alignment enhancement processing on the mid-to-deep features. After semantic alignment enhancement processing, the contrast between the calcified strong echo region and the sound shadow region is focused, and a mid-to-deep semantic alignment enhancement feature map is generated. The spatial orientation capture process of the mid-to-deep semantic alignment enhancement feature map is performed by the directional dependency modeling unit to capture the directional propagation features of the calcified strong echo region and the sound shadow region, and to generate the sound shadow basic feature map. The feature dimension of the sound shadow basic feature map is consistent with the feature dimension of the mid-to-deep features. The audio-visual post-processing module performs at least feature processing on the basic audio-visual feature map to generate the audio-visual target feature map after feature processing.

10. The intravascular ultrasound lithotripsy system integrating IVUS and IVL according to claim 9, characterized in that, The directional dependency modeling unit may include a difference mapping module, a direction-aware module, and a radial spatial attention module connected in sequence, wherein... The difference mapping module performs feature filtering and difference mapping on the mid-to-deep semantic alignment enhancement feature map, and generates a strong echo-low echo difference feature map. The feature filtering and difference mapping process includes two branches. In one branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a sigmoid function operation. In the other branch, a 1×1 convolution operation is performed on the mid-to-deep semantic alignment enhancement feature map, followed by a tanh operation. After that, the results of the two branches are multiplied point by point. After generating the strong echo-low echo difference feature map, the direction perception module performs direction perception processing on the strong echo-low echo difference feature map so that a direction perception feature map can be generated after the direction perception processing. In the direction perception processing, at least a 3×3 convolution with angle weights is used to perform convolution processing on the strong echo-low echo difference feature map. After generating the orientation-aware feature map, the radial spatial attention module first performs radial spatial attention processing to generate a radial spatial attention feature map. Then, the radial spatial attention feature map and the orientation-aware feature map are multiplied point by point to generate the basic sound and shadow feature map after the point-by-point multiplication.

Images