Image quality evaluation method, device and computer readable storage medium

Abstract

The application discloses an image quality evaluation method, comprising: acquiring a plurality of two-dimensional medical images of a target object; determining at least one image quality index of each two-dimensional medical image based on a first evaluation model generated by training; the image quality index is used to represent an influencing factor of quality evaluation of the target object; determining an image quality evaluation result of a three-dimensional medical image of the target object based on a second evaluation model generated by training and the at least one image quality index of each two-dimensional medical image; the image quality evaluation result comprises an evaluation index used to represent the completeness of the target object. The application also discloses an image quality evaluation device and a computer readable storage medium. The application can realize accurate and comprehensive evaluation of image quality, and can be used to predict the image segmentation effect of the target object.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and in particular to an image quality assessment method, apparatus, and computer-readable storage medium. Background Technology

[0002] In natural cavity robotic surgery, preoperative planning requires segmentation of target tissues, such as lung CT images. However, the image segmentation results are often affected by the quality of the input CT images; only with higher quality input CT images can the bronchial segmentation results meet the requirements. Therefore, quality assessment of CT images is of great significance. However, how to accurately and comprehensively assess image quality remains an area of ​​research. Summary of the Invention

[0003] The main objective of this application is to provide an image quality assessment method, apparatus, and computer-readable storage medium, aiming to solve the technical problem that existing image quality assessment methods are not accurate and comprehensive enough.

[0004] To achieve the above objectives, this application provides an image quality assessment method, the method comprising the following steps: Acquire multiple two-dimensional medical images of the target object; At least one image quality index is determined for each of the two-dimensional medical images based on a first evaluation model generated through training; the image quality index is used to characterize the factors influencing the quality assessment of the target object; Based on a trained second evaluation model and at least one image quality index for each of the two-dimensional medical images, an image quality evaluation result for the three-dimensional medical image of the target object is determined; the image quality evaluation result includes an evaluation index used to characterize the integrity of the target object.

[0005] Optionally, the image quality metrics include at least one of noise level, artifact level, lung consolidation level, and inflammation level.

[0006] Optionally, the image quality metric may also include visibility.

[0007] Optionally, the visibility is obtained by fusing the noise level, the degree of artifact, the degree of lung consolidation, and the degree of inflammation using the first evaluation model.

[0008] Optionally, the image quality indicators include at least the degree of lung consolidation and the degree of inflammation.

[0009] Optionally, the image quality assessment result further includes a quality description; determining the image quality assessment result of the three-dimensional medical image of the target object based on the trained second assessment model and at least one image quality index for each of the two-dimensional medical images includes: Based on at least one image quality index for each of the two-dimensional medical images, the target object is evaluated using a trained second evaluation model to obtain an evaluation index that characterizes the integrity of the target object. Based on the frequency of occurrence of different values ​​or intervals of each of the image quality indicators, and the pre-defined correspondence between different values ​​or intervals of the image quality indicators and the quality description, the quality description of the three-dimensional medical image of the target object is determined.

[0010] Optionally, the step of evaluating the image quality of the target object using a trained second evaluation model based on at least one image quality index for each of the two-dimensional medical images to obtain an evaluation index characterizing the integrity of the target object includes: Obtain the slice thickness value of the two-dimensional medical image; At least one image quality index and the slice thickness value of each of the two-dimensional medical images are input into a second evaluation model that has been trained to obtain an evaluation index predicted by the second evaluation model to characterize the integrity of the target object.

[0011] Optionally, determining the quality description of the three-dimensional medical image of the target object based on the frequency of occurrence of different values ​​or intervals of each of the statistically defined image quality indicators, and the preset correspondence between different values ​​or intervals of the image quality indicators and the quality description, includes: Based on the frequency of occurrence of different values ​​or intervals of the target image quality index, at least one target value or target interval of the target image quality index is determined; the target image quality index is any image quality index; the target value or the target interval is a value or interval that satisfies a preset condition and has a frequency greater than a preset frequency threshold. Based on the pre-defined correspondence between different values ​​or intervals of the target image quality index and the quality description, determine the quality description corresponding to the target value or the target interval. A quality description of the three-dimensional medical image of the target object is determined based on the quality description corresponding to the target value or the target interval value.

[0012] Optionally, the first evaluation model is obtained by training based on historical two-dimensional medical images and corresponding historical image quality indicators; The second evaluation model is obtained by training based on historical image quality indicators and corresponding historical evaluation indicators corresponding to historical two-dimensional medical images.

[0013] This application also provides an image quality assessment device, including: a memory and a processor, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, it implements the steps of the method described above.

[0014] Optionally, the device further includes a display configured to display image quality assessment results of a three-dimensional medical image of the target object.

[0015] Optionally, the display is further configured to display at least one of the following: the number of two-dimensional medical images of the target object, the slice thickness value, and the image quality index.

[0016] This application also provides a computer-readable storage medium storing computer program instructions configured to be loaded by a processor and executed to implement the steps of the method described above.

[0017] The image quality assessment method, apparatus, and computer-readable storage medium provided in this application, after obtaining at least one image quality index for each two-dimensional medical image of the target object through a first assessment model, obtain the image quality assessment result of the three-dimensional medical image of the target object based on the at least one image quality index of each two-dimensional medical image through a second assessment model. Thus, by effectively integrating image quality indices from different dimensions to assess image quality, accurate and comprehensive image quality assessment can be achieved, which can then be used to predict the image segmentation effect on the target object. Attached Figure Description

[0018] Figure 1 A schematic diagram of a catheter system provided in one embodiment of this application is shown.

[0019] Figure 2 A schematic diagram of a catheter device and power unit provided in an embodiment of this application is shown.

[0020] Figure 3 A schematic diagram of a catheter device provided in one embodiment of this application is shown.

[0021] Figure 4 A schematic flowchart of an image quality assessment method provided in an embodiment of this application is shown.

[0022] Figure 5 This paper presents a 3D morphological diagram of the bronchi corresponding to different rating levels provided in an embodiment of this application.

[0023] Figure 6 A 2D schematic diagram of a bronchus with different levels of visibility is shown in one embodiment of this application.

[0024] Figure 7 A schematic diagram of the structure of a first evaluation model provided in an embodiment of this application is shown.

[0025] Figure 8 A schematic diagram of the structure of a second evaluation model provided in an embodiment of this application is shown.

[0026] Figure 9 A schematic diagram of the structure of an image quality assessment device provided in an embodiment of this application is shown.

[0027] Figure 10 A schematic diagram of a computer-readable storage medium provided in one embodiment of this application is shown.

[0028] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0030] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0031] In this application, unless otherwise expressly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0032] Furthermore, the use of terms such as "first" and "second" in this application is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. If the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed in this application.

[0033] Figure 1 An embodiment of the catheter system 1000 provided in this application is illustrated. The catheter system 1000 includes an imaging cart 100, a trolley 200 connected to the imaging cart 100, a user input device 300, a catheter instrument 400 that can be attached to the trolley 200, a sensor system 500 connected to the trolley 200, and a control system 600 for controlling the catheter instrument 400, the user input device 300, the sensor system 500, and the imaging cart 100. The user input device 300 can be wired or wirelessly connected to the trolley 200. When an operator performs various procedures on a patient near the trolley 200, they can trigger control commands by operating the user input device 300, which, driven by the trolley 200, controls the catheter instrument 400 to move forward, retract, and bend / turn.

[0034] The trolley 200 can typically be moved to the side of the operating table to engage the catheter instrument 400. Under control commands, it controls the catheter instrument 400 to move vertically, horizontally, or in non-vertical and non-horizontal directions, thus providing a better preoperative preparation angle for the operation of the catheter instrument 400. The control commands can be triggered by the operator through the user input device 300, or by the operator directly clicking or pressing buttons on the trolley 200. In other embodiments, the control commands can also be voice control or commands triggered through a force feedback mechanism.

[0035] like Figure 1As shown, the trolley 200 may further include a base 210, a sliding seat 220 that can move up and down along the base 210, and two robotic arms 230 fixedly connected to the sliding seat 220. Each robotic arm 230 may include multiple arm segments connected at joints, providing multiple degrees of freedom, for example, seven degrees of freedom corresponding to seven arm segments. A power unit (not shown) is installed at the end of each robotic arm 230. The power unit engages the catheter instrument 400 and, under the driving action of the power unit, controls the end of the catheter instrument 400 to bend and turn accordingly. The two robotic arms 230 may have identical or partially identical structures; one robotic arm 230 engages the internal catheter instrument 410, and the other robotic arm 230 engages the external catheter instrument 420. During installation, the external catheter device 420 can be installed first. After the external catheter device 420 is installed, the catheter of the internal catheter device 410 is inserted into the catheter of the external catheter device 420.

[0036] The sensor system 500 has one or more subsystems for receiving information about the catheter device 400. The subsystems may include: a position sensor system; a shape sensor system for determining the position, orientation, velocity, rate, pose, and / or shape of the distal end of the catheter device 400 and / or along one or more segments of the catheter that may constitute the catheter device 400; and / or a visualization system for capturing images from a camera at the distal end of the catheter device 400.

[0037] The imaging vehicle 100 may be equipped with a display system 110 and a flushing system (not shown in the figure), etc. The display system 110 is used to display images or representations of the surgical site and catheter instruments 400 generated by the subsystems of the sensor system 500. It can also display real-time images of the surgical site and catheter instruments 400 captured by a visualization system. Image data from imaging technologies such as computed tomography (CT), magnetic resonance imaging (MRI), optical coherence tomography (OCT), and ultrasound can also be used to present images of the surgical site recorded preoperatively or intraoperatively. Preoperative or intraoperative image data can be presented as two-dimensional, three-dimensional, or four-dimensional (e.g., time-based or rate-based information) images and / or as images from models created based on preoperative or intraoperative image datasets, and virtual navigation images can also be displayed. In the virtual navigation images, the actual position of the catheter instruments 400 is registered with the preoperative images to present a virtual image of the catheter instruments 400 within the surgical site to the operator from the outside.

[0038] The control system 600 includes at least one memory and at least one processor communicatively connected to the robotic arm. It is understood that the control system 600 can be integrated into the trolley 200 or the imaging cart 100, or it can be set up independently. The control system 600 can support wireless communication protocols such as IEEE 802.11, IrDA, Bluetooth, HomeRF, DECT, and wireless telemetry. The control system 600 can transmit one or more signals instructing the catheter device 400 to move, which is then moved by the power unit. The catheter device 400 can extend to the surgical site within the body via an opening in the patient's natural cavity or a surgical incision.

[0039] Furthermore, the control system 600 may include a mechanical control system (not shown in the figure) and an image processing system (not shown in the figure). The mechanical control system is used to control the movement of the catheter instrument 400, and therefore can be integrated into the trolley 200. The image processing system is used for virtual navigation path planning, and therefore can be integrated into the imaging vehicle 100. Of course, the various subsystems of the control system 600 are not limited to the specific cases listed above, and can be reasonably set according to actual conditions. Among them, the image processing system can image the surgical site based on images of the surgical site recorded before or during the operation, using the above-mentioned imaging technology. Software that can be used in conjunction with manual input can also convert the recorded images into two-dimensional or three-dimensional synthetic images of parts or the entire anatomical organ or segment. During the virtual navigation procedure, the sensor system 500 can be used to calculate the position of the catheter instrument 400 relative to the patient's anatomical structure. This position can be used to generate external tracking images and internal virtual images of the patient's anatomical structure, realizing the registration of the actual position of the catheter instrument 400 with the preoperative image, thereby presenting a virtual image of the catheter instrument 400 within the surgical site to the operator from the outside.

[0040] The internal catheter device 410 and the external catheter device 420 have largely the same structure, each having a slender and flexible internal catheter 41 and an external catheter 42, respectively. The diameter of the external catheter 42 is slightly larger than that of the internal catheter 41, so that the internal catheter 41 can pass through the lumen of the external catheter 42 and provide some support for the internal catheter 41. This allows the internal catheter 41 to reach the target location in the patient's body, so as to facilitate operations such as tissue or cell sampling from the target location.

[0041] Certain movements of the user input device 300 can cause corresponding movements of the catheter instrument 400. For example, when the operator moves the direction lever of the user input device 300 up or down, the movement of the direction lever can be mapped to a corresponding pitch movement of the end of the catheter instrument 400; when the operator moves the direction lever of the user input device 300 left or right, the movement of the direction lever can be mapped to a corresponding yaw movement of the end of the catheter instrument 400. In this embodiment, the user input device 300 can control the end of the catheter instrument 400 to move within a 360° spatial range.

[0042] Figure 2 and Figure 3 An embodiment of the present application illustrates a catheter device 400. The catheter device 400 is configured to engage with a power unit 240 of a robotic arm 230. The catheter device 400 includes an instrument housing 45 configured to engage with the power unit 240 and a catheter 48 connected to the instrument housing 45. "Engagement" refers to a state where, when the instrument housing 45 is installed in the power unit 240, the driving force of the power unit 240 can be transmitted to the instrument housing 45, enabling the catheter 48 to move normally. For example, under the driving force of the power unit 240, the end of the catheter 48 can bend or change direction. The catheter 48 may include the aforementioned inner catheter 41 and outer catheter 42. In this embodiment, unless otherwise specified, the catheter is not distinguished as an inner or outer catheter.

[0043] The instrument box 45 includes multiple drive wheels 451 configured to be driven by a power unit 240 and multiple drive wires 452. The power unit 240 includes multiple drive motors 241, and the drive motors 241, drive wheels 451, and drive wires 452 are arranged in a one-to-one correspondence. Each drive wheel 451 is configured to engage with a corresponding drive motor 241. That is, when the instrument box 45 is installed on the power unit 240, the corresponding drive motor 241 can drive the drive wheel 451 to rotate, and the corresponding drive wire 452 is wound on the drive wheel 451. The movable part of the corresponding drive wire 452, that is, the part not wound on the drive wheel 451, extends into the catheter 48, extends along the length of the catheter 48, and is finally fixed to the end of the catheter.

[0044] In this application, the end, also referred to as the distal end or head, refers to the end away from the instrument box 45; the anterior end, also referred to as the proximal end or tail, refers to the end close to the instrument box 45.

[0045] A portion of the conduit 48, including its distal end, is a controllable segment 49, the distal end of which is the distal end of the conduit 48. The controllable segment 49 can be a joint assembly with high stiffness in the extension direction and low stiffness in the bending direction, capable of bending under the control of the drive wire 452, thereby achieving the steering of the conduit 48. In some embodiments, this joint assembly may be referred to as a snake skeleton.

[0046] In natural cavity robotic surgery, preoperative planning requires segmentation of target tissues, such as lung CT images. However, the image segmentation results are often affected by the quality of the input CT images; only with higher quality input CT images can the bronchial segmentation results meet the requirements. Therefore, quality assessment of CT images is of great significance.

[0047] The processor of the control system 600 is configured to perform the following steps to implement the image quality assessment method provided in the first embodiment of this application. For example... Figure 4 As shown, this method includes: Step S10: Obtain multiple two-dimensional medical images of the target object.

[0048] The target object can be any object for which image quality assessment is to be performed, and can be human physiological tissue, including but not limited to bronchi and blood vessels, etc., without specific limitations. Preferably, acquiring multiple two-dimensional medical images of the target object can be done by selecting multiple two-dimensional medical images of the middle region from a set of two-dimensional medical images obtained by sequentially scanning the target object.

[0049] Step S20: Determine at least one image quality index for each two-dimensional medical image based on the first evaluation model generated by training; the image quality index is used to characterize the influencing factors of the quality assessment of the target object.

[0050] Specifically, each two-dimensional medical image is sequentially input into a first evaluation model that has been trained, and at least one image quality index is obtained for each two-dimensional medical image predicted and output by the first evaluation model that has been trained.

[0051] The first evaluation model is obtained by training on historical two-dimensional medical images and corresponding historical image quality indicators. It can be understood that historical two-dimensional medical images of different patients and their corresponding historical image quality indicators can be obtained as training samples to train the generated first evaluation model. The first evaluation model can be constructed based on artificial intelligence algorithms such as neural networks and genetic algorithms. The training process of the first evaluation model can be found in existing technologies and will not be elaborated upon here.

[0052] In some embodiments, image quality metrics include at least one of noise level, artifact level, lung consolidation level, and inflammation level. Noise level indicates the intensity fluctuation of random signals in the image; low noise levels result in clear images with easily distinguishable details (such as microvessels and small nodules), while high noise levels result in blurred images that may mask lesions (such as ground-glass opacities). Artifact level indicates interference signals from non-anatomical structures in the image, which may mislead diagnosis. Lung consolidation level indicates the degree of density increase caused by fluid or cell filling of lung tissue. Inflammation level indicates the strength of the tissue's immune response to damage or infection. Noise level and artifact level characterize image quality, while lung consolidation level and inflammation level characterize pathological features. Thus, by acquiring image quality metrics across different dimensions, the accuracy and comprehensiveness of image quality assessment can be improved.

[0053] In some embodiments, the image quality metric further includes visibility. Visibility is used to characterize the ease of extracting the target object. It can be understood that since visibility characterizes the ease of extracting the target object, i.e., the ease of segmentation, the quality of the target object segmentation result for each two-dimensional medical image can be determined by the visibility of that image.

[0054] In some embodiments, visibility is obtained by fusing noise level, artifact level, lung consolidation level, and inflammation level using a first evaluation model.

[0055] Since the segmentation result of the target object is mainly affected by imaging quality and image content, and these effects can be represented by four dimensions: noise level, artifact level, lung solidification level, and inflammation level, in order to obtain visibility to characterize the ease of extraction of the target object, a first evaluation model can be used to fuse and calculate the noise level, artifact level, lung solidification level, and inflammation level to determine the visibility of the two-dimensional medical image. Thus, by learning the correlation between the features extracted from the four dimensions of noise level, artifact level, lung solidification level, and inflammation level and visibility, multiple dimensions of image quality indicators can be effectively integrated for image quality evaluation, thereby improving the accuracy of image quality assessment.

[0056] In some embodiments, image quality metrics include at least the degree of lung solidification and the degree of inflammation. It is understood that the degree of lung solidification and the degree of inflammation can affect the clarity of the target object; in cases where there is a large amount of solidified material, the target object may not be visible. It should be noted that in this embodiment, label values ​​such as 1, 2, 3, etc., can be used to represent the values ​​or ranges of image quality metrics.

[0057] Step S30: Based on the trained second evaluation model and at least one image quality index for each two-dimensional medical image, determine the image quality evaluation result of the three-dimensional medical image of the target object; the image quality evaluation result includes evaluation indices used to characterize the integrity of the target object.

[0058] Specifically, at least one image quality index of all two-dimensional medical images is input into a trained second evaluation model to obtain the image quality evaluation result of the three-dimensional medical image of the target object predicted by the trained second evaluation model.

[0059] The second evaluation model is obtained by training on historical image quality indicators and corresponding historical evaluation indicators based on historical two-dimensional medical images. In other words, historical image quality indicators and historical evaluation indicators corresponding to historical two-dimensional medical images of different patients can be obtained as training samples to train the generated second evaluation model. The second evaluation model can be constructed based on artificial intelligence algorithms such as neural networks and genetic algorithms. The training process of the second evaluation model can be found in existing technologies and will not be elaborated upon here.

[0060] The evaluation indicators can be specific to scores, for example, a higher score indicates a higher degree of completeness of the target object, while a lower score indicates a lower degree of completeness of the target object.

[0061] In summary, the image quality assessment method provided in the above embodiments, after obtaining at least one image quality index for each two-dimensional medical image of the target object through a first assessment model, obtains the image quality assessment result of the three-dimensional medical image of the target object based on the at least one image quality index of each two-dimensional medical image through a second assessment model. Thus, by effectively integrating image quality indices from different dimensions to assess image quality, accurate and comprehensive image quality assessment can be achieved, which can then be used to predict the image segmentation effect on the target object.

[0062] In some embodiments, step S30 includes: Obtain the slice thickness value of a two-dimensional medical image; At least one image quality index and slice thickness value of each two-dimensional medical image are input into a trained second evaluation model to obtain an evaluation index predicted by the second evaluation model to characterize the integrity of the target object.

[0063] In this model, the slice thickness value of each 2D medical image of the target object can be considered the same. Slice thickness represents the physical thickness of a single image layer (in millimeters), directly affecting the image's spatial resolution, signal-to-noise ratio, and clinical diagnostic capability. Considering the significant impact of slice thickness on the target object, it can be used as a factor influencing the image quality assessment of the 3D medical image of the target object, i.e., as an input variable to the second evaluation model. Specifically, at least one image quality index and the slice thickness value of all 2D medical images are input into the trained second evaluation model to obtain the predicted image quality assessment result of the 3D medical image of the target object. This further improves the accuracy and comprehensiveness of image quality assessment, thereby further enhancing the prediction accuracy of image segmentation results for the target object.

[0064] In some embodiments, the image quality assessment result further includes a quality description; step S30 includes: Based on at least one image quality index for each two-dimensional medical image, the image quality of the target object is evaluated using a trained second evaluation model to obtain an evaluation index that characterizes the integrity of the target object. Based on the frequency of occurrence of different values ​​or intervals of each image quality index in the statistics, and the correspondence between the different values ​​or intervals of the preset image quality index and the quality description, the quality description of the three-dimensional medical image of the target object is determined.

[0065] Since each two-dimensional medical image corresponds to at least one image quality index, and each image quality index corresponds to a value or range of values, the frequency of different values ​​or ranges of each image quality index can be statistically analyzed based on at least one image quality index of all two-dimensional medical images. For example, taking 200 two-dimensional medical images with image quality indicators including noise level and target object as an example, assuming the noise level ranges from 1 to 5, the smaller the noise level value, the more severe the noise impact, and the larger the noise level value, the less the noise impact. The noise level values ​​corresponding to different two-dimensional medical images may be different. By statistically analyzing the frequency of different noise level values ​​for all two-dimensional medical images, we can find that there are 10 two-dimensional medical images with a noise level value of 1, 20 with a noise level value of 2, 20 with a noise level value of 3, 100 with a noise level value of 4, and 150 with a noise level value of 5. That is, the noise level value of 1 appears 10 times, the noise level value of 2 appears 20 times, the noise level value of 3 appears 20 times, the noise level value of 4 appears 100 times, and the noise level value of 5 appears 150 times.

[0066] The quality description is used to describe the image characteristics of the 3D medical image of the target object. After statistically analyzing the frequency of different values ​​or ranges of each image quality index, the quality description of the 3D medical image of the target object can be determined by combining the preset correspondence between different values ​​or ranges of the image quality index and the quality description. Specifically, for each image quality index, the quality description corresponding to the most frequent value or range can be used as the quality description of the 3D medical image of the target object, or the quality description of the 3D medical image of the target object can be determined based on the quality descriptions corresponding to multiple most frequent values ​​or ranges.

[0067] Thus, compared to existing image quality assessment methods that typically only have a total score, which suffers from poor interpretability and is difficult for users to understand, this application not only includes assessment indicators to characterize the integrity of the target object, but also quality descriptions, improving the interpretability of image quality assessment results and making them easier for users to understand.

[0068] In some embodiments, the quality description of the three-dimensional medical image of the target object is determined based on the frequency of occurrence of different values ​​or intervals of each statistically analyzed image quality index, and the pre-defined correspondence between different values ​​or intervals of the image quality index and the quality description, including: Based on the frequency of occurrence of different values ​​or intervals of the target image quality index, at least one target value or target interval of the target image quality index is determined; the target image quality index can be any image quality index; the target value or target interval is a value or interval that meets the preset conditions and has a frequency greater than the preset frequency threshold. Based on the correspondence between different values ​​or intervals of the preset target image quality index and the quality description, determine the quality description corresponding to the target value or target interval. The quality description of the three-dimensional medical image of the target object is determined based on the quality description corresponding to the target value or target interval value.

[0069] By statistically analyzing the frequency of different values ​​or ranges for each image quality indicator, at least one target value or target range for the target image quality indicator can be determined. Preset conditions can be set according to actual needs. For example, if a larger image quality indicator value or a larger minimum value within a range indicates better image quality, the preset condition could be a value less than a preset first threshold or a maximum value within a range less than a preset second threshold. Conversely, if a smaller image quality indicator value or a smaller maximum value within a range indicates better image quality, the preset condition could be a value greater than a preset first threshold or a minimum value within a range greater than a preset second threshold, and so on.

[0070] For a target image quality index, there may be only one target value or multiple target range values. If there are multiple target values ​​or target range values ​​for a target image quality index, only the quality description corresponding to the most frequent target value or target range value needs to be obtained. Taking the bronchus as the target object and visibility as an image quality indicator, let's assume that the visibility value range can include 1, 2, 3, 4, and 5. The quality description corresponding to a visibility value of 1 is "approximately half of the lung is solidified," the quality description corresponding to a visibility value of 2 is "approximately one-quarter of the lung is solidified," the quality description corresponding to a visibility value of 3 is "there are relatively serious inflammation or global image problems such as noise and artifacts," the quality description corresponding to a visibility value of 4 is "there are mild noise, artifacts, or inflammatory lesions," and the quality description corresponding to a visibility value of 5 is "the lung is normal or only has local nodular lesions." If, based on the frequency of different visibility values, at least one target value for visibility is determined to include 4 and 5, and the frequency of visibility value 4 is less than the frequency of visibility value 5, then the quality description corresponding to the target visibility value of 5 can be determined as the quality description of visibility.

[0071] In this process, after obtaining the quality descriptions corresponding to the target values ​​or target intervals of all target image quality indicators, these descriptions can be used to define the quality descriptions of the three-dimensional medical images of the target object. This allows for the presentation of easily understandable quality descriptions to the user, thus enabling the interpretation of the three-dimensional medical images of the target object. Furthermore, by considering the frequency of different values ​​or intervals of each image quality indicator and the pre-defined correspondence between different values ​​or intervals of the image quality indicators and their quality descriptions, accurate quality descriptions of the three-dimensional medical images of the target object can be determined, further improving the interpretability of the image quality assessment results.

[0072] Based on the same inventive concept as the foregoing embodiments, the foregoing embodiments will be described in detail below through a specific example. In this example, the target object is the bronchus, the medical image is a CT image (also known as CT data), the two-dimensional medical image is a 2D slice, and the evaluation index is a score.

[0073] The image quality assessment method provided in this embodiment mainly consists of the following steps: 1. Scoring of the bronchi For each bronchial CT image, its integrity was scored to serve as a label for training samples. The scoring was divided into five levels, and the descriptions corresponding to the five levels are shown in Table 1. The 3D morphology of the bronchus corresponding to each level is shown in Table 1. Figure 5 As shown, Figure 5 (1) in the figure represents the 3D morphology of the bronchus corresponding to the rating level 1, and so on.

[0074] Table 1

[0075] 2. The 2D slices (also known as 2D images) of CT data are scored from multiple dimensions. The specific steps are as follows: (1) Four variables with high acceptance were selected: noise level, artifact degree, lung consolidation degree, and inflammation degree, and were scored in terms of imaging quality and image content respectively.

[0076] (2) The visibility of the bronchi is selected as the total score of the 2D image. This index is scored by observing whether the bronchial features are obvious in the 2D slice. The score can reflect the difficulty of extracting bronchial features.

[0077] (3) For each 2D slice of the CT data, a score is assigned using the scoring items described in (1) and (2) above, with each scoring item having a score ranging from 1 to 5 points. For example, Table 2 shows the scoring criteria for the visibility of the bronchi, and the corresponding 2D image morphology is as follows: Figure 6 As shown.

[0078] Table 2

[0079] 3. Deep learning methods are used to fit human scoring rules to train the first evaluation model. Prior knowledge indicates that bronchial segmentation results are primarily affected by image quality and image content, which are represented by four dimensions: noise level, artifact level, lung consolidation level, and inflammation level. In the first evaluation model, these four dimensions are supervised separately, and fused features are used to obtain a 2D image score. (See also...) Figure 7 The diagram below illustrates the structure of the first evaluation model. The 2D slices used for training are taken as input, and N features are extracted through the first evaluation model (N is 4 in this embodiment). The corresponding predicted values ​​N (i.e., the predicted values ​​of noise level, artifact degree, lung consolidation degree, and inflammation degree) are calculated respectively. The N features are then fused to obtain the predicted 2D image score. Finally, the corresponding loss is calculated by combining the noise level, artifact degree, lung consolidation degree, inflammation degree, and 2D image score of the 2D slice annotations. The model parameters are then adjusted based on the loss to train the model.

[0080] 4. Key slices of the CT data are scored, and through statistics and classification, the score of each 2D image and the overall score of each dimension are obtained. The architecture of the second evaluation model for image quality assessment is as follows: Figure 8 As shown.

[0081] Specifically, n slices from the middle three-quarters region of the CT data along the Z-axis are sequentially input into the first evaluation model to obtain a comprehensive score and scores for each slice across four dimensions (i.e., components). The frequency of each score in the CT data (i.e., volume data) is statistically analyzed, and a vector of length 5 (f1, f2, f3, f4, f5) is constructed from these frequencies, where each element represents a score for one dimension. Simultaneously, based on prior knowledge that slice thickness significantly affects the bronchi, slice thickness information is transformed and used as a feature, forming a feature set of length 6. Then, a classifier (i.e., the second evaluation model) is established to correlate these features with the final total score of the CT data.

[0082] 5. Interpret the scores. Through the preceding steps, we obtain the total score of the entire CT data and the frequency information of scores in different dimensions. Now, we need to describe the image characteristics based on this information. To address this issue, this embodiment proposes using features that reach a low score threshold as descriptive keywords, combining all descriptive keywords to form a general description. For example, if the lung consolidation score is 1 and its frequency is relatively high, then the descriptive term "large lung consolidation area" is derived; if the noise level score is 1 and its frequency is relatively high, then the descriptive term "high noise" is derived; and if the other items do not have low-score features, then no description is made. Therefore, the general description is: "Large lung consolidation area and high noise." For example, for the image quality index value (i.e., score), the lower the score, the more serious the problem. Assuming the noise level is rated from 1 to 5, a score of 1 indicates severe noise impact, and a score of 5 indicates almost no noise impact. If a 3D medical image of a bronchus consists of 300 slices of 2D medical images, and 100 of these 300 2D medical images have a noise level score of less than 2, while the threshold is 50, then the data has a significant noise impact, and corresponding descriptive terms need to be generated.

[0083] In summary, the image quality assessment method provided in the above embodiments, on the one hand, predicts the segmentation of the bronchi based on the total score and scores of each dimension, and generates descriptive terms for the overall CT data based on the score features of each dimension, making it easier for users to understand. On the other hand, a "multi-branch network" is designed, which can score each influencing factor specifically, and then integrate the features of this branch network to achieve a comprehensive score, effectively integrating the indicators of each dimension. Furthermore, a scoring standard for CT images is proposed, which scores 2D slice images from four dimensions and a comprehensive score that reflects the difficulty of bronchial segmentation, allowing the model to learn the relationship between the features extracted from different scoring dimensions and the comprehensive score. Moreover, when scoring the CT image as a whole, only the segmentation results are used; therefore, the overall CT score can reflect the bronchial segmentation effect and can be used to predict the quality of bronchial segmentation results.

[0084] This application also provides an image quality assessment device. Please refer to... Figure 9 This illustrates a schematic diagram of the structure of an image quality assessment device provided in an embodiment of this application. Figure 9 As shown, the image quality assessment device includes: a processor 60, a memory 61, a bus 62, and a communication interface 63. The processor 60, the communication interface 63, and the memory 61 are connected via the bus 62. The memory 61 stores computer program instructions that can be executed by the processor 60. When the processor 60 executes the computer program instructions, it executes the image quality assessment method provided in any of the foregoing embodiments of this application.

[0085] The memory 61 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 63 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0086] Bus 62 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 61 is used to store programs. After receiving an execution instruction, the processor 60 executes the program. The image quality assessment method disclosed in any of the foregoing embodiments of this application can be applied to the processor 60, or implemented by the processor 60.

[0087] Processor 60 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 60 or by instructions in software form. Processor 60 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), an Off-the-shelf Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 61. Processor 60 reads the information in memory 61 and, in conjunction with its hardware, completes the steps of the above method.

[0088] The image quality assessment device and the image quality assessment method provided in this application are based on the same application concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0089] In some embodiments, the device further includes a display 64 configured to display image quality assessment results of a three-dimensional medical image of the target object.

[0090] In some embodiments, the display 64 is further configured to display at least one of the following: the number of two-dimensional medical images of the target object, the slice thickness value, and the image quality index.

[0091] This application also provides a computer-readable storage medium corresponding to the image quality assessment method provided in the foregoing embodiments. Please refer to... Figure 10 The computer-readable storage medium 6 shown thereon stores computer program instructions that, when executed by a processor, implement the image quality assessment method provided in any of the foregoing embodiments.

[0092] It should be noted that examples of the computer-readable storage medium may include, but are not limited to, optical discs, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other optical and magnetic storage media, which will not be elaborated here.

[0093] The computer-readable storage medium provided in the above embodiments of this application and the image quality assessment method provided in the embodiments of this application are based on the same application concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0094] The above description is merely an optional embodiment of this application and does not limit the patent scope of this application. Any equivalent structural transformations made based on the content of the specification and drawings of this application under the concept of this application, or direct / indirect applications in other related technical fields, are included within the patent protection scope of this application.

Claims

1. An image quality assessment method, characterized in that, The method includes the following steps: Acquire multiple two-dimensional medical images of the target object; At least one image quality index is determined for each of the two-dimensional medical images based on a first evaluation model generated through training; the image quality index is used to characterize the factors influencing the quality assessment of the target object; Based on a trained second evaluation model and at least one image quality index for each of the two-dimensional medical images, an image quality evaluation result for the three-dimensional medical image of the target object is determined; the image quality evaluation result includes an evaluation index used to characterize the integrity of the target object.

2. The method as described in claim 1, characterized in that, The image quality metrics include at least one of the following: noise level, artifact level, lung consolidation level, and inflammation level.

3. The method as described in claim 2, characterized in that, The image quality metrics also include visibility.

4. The method as described in claim 3, characterized in that, The visibility level is obtained by fusing the noise level, the degree of artifact, the degree of lung solidification, and the degree of inflammation using the first evaluation model.

5. The method as described in claim 1, characterized in that, The image quality indicators include at least the degree of lung consolidation and the degree of inflammation.

6. The method according to any one of claims 1 to 4, characterized in that, The image quality assessment result also includes a quality description; the determination of the image quality assessment result of the three-dimensional medical image of the target object based on the trained second assessment model and at least one image quality index for each of the two-dimensional medical images includes: Based on at least one image quality index for each of the two-dimensional medical images, the target object is evaluated using a trained second evaluation model to obtain an evaluation index that characterizes the integrity of the target object. Based on the frequency of occurrence of different values ​​or intervals of each of the image quality indicators, and the pre-defined correspondence between different values ​​or intervals of the image quality indicators and the quality description, the quality description of the three-dimensional medical image of the target object is determined.

7. The method as described in claim 6, characterized in that, The image quality assessment of the target object is performed using a trained second evaluation model based on at least one image quality index for each of the two-dimensional medical images, to obtain an evaluation index characterizing the integrity of the target object, including: Obtain the slice thickness value of the two-dimensional medical image; At least one image quality index and the slice thickness value of each of the two-dimensional medical images are input into a second evaluation model that has been trained to obtain an evaluation index predicted by the second evaluation model to characterize the integrity of the target object.

8. The method as described in claim 6, characterized in that, The step of determining the quality description of the three-dimensional medical image of the target object based on the frequency of occurrence of different values ​​or intervals of each of the image quality indicators and the preset correspondence between different values ​​or intervals of the image quality indicators and the quality description includes: Based on the frequency of occurrence of different values ​​or intervals of the target image quality index, at least one target value or target interval of the target image quality index is determined; the target image quality index is any image quality index; the target value or the target interval is a value or interval that satisfies a preset condition and has a frequency greater than a preset frequency threshold. Based on the pre-defined correspondence between different values ​​or intervals of the target image quality index and the quality description, determine the quality description corresponding to the target value or the target interval. A quality description of the three-dimensional medical image of the target object is determined based on the quality description corresponding to the target value or the target interval value.

9. The method as described in claim 6, characterized in that, The first evaluation model was obtained by training based on historical two-dimensional medical images and corresponding historical image quality indicators. The second evaluation model is obtained by training based on historical image quality indicators and corresponding historical evaluation indicators corresponding to historical two-dimensional medical images.

10. An image quality assessment device, comprising: A memory and a processor, wherein the memory stores computer program instructions for execution on the processor, and the processor, when executing the computer program instructions, implements the steps of the method as described in claims 1 to 9.

11. The apparatus as claimed in claim 10, characterized in that, The device also includes a display configured to display image quality assessment results of a three-dimensional medical image of the target object.

12. The apparatus as claimed in claim 11, characterized in that, The display is also configured to display at least one of the following: the number of two-dimensional medical images of the target object, the slice thickness value, and the image quality index.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions configured to be loaded by a processor and executed to implement the steps of the method as described in claims 1 to 9.

Images