Ultrasound imaging systems and ultrasound image analysis methods

Abstract

An ultrasound imaging system and ultrasound image analysis method are disclosed. The ultrasound imaging system includes an ultrasound probe, a transmitting circuit, a receiving circuit, a processor, and a display. The processor is used to: acquire ultrasound images of the thyroid region of a subject; detect lesions within the image and identify TI-RADS lesion features corresponding to at least five TI-RADS feature types; determine a TI-RADS score and thereby determine the TI-RADS classification of the lesion; generate a radar chart using the TI-RADS feature type as the classification axis; wherein the classification axis divides the radar chart into multiple partitions, each classification axis or each partition representing a TI-RADS feature type, and at least one classification axis or partition has a scale unit for representing the score; generate a feature graph on the radar chart based on the TI-RADS score; and the display is used to display the radar chart, feature graph, and TI-RADS classification. This invention uses radar charts and feature graphs to intuitively present the lesion score, which is beneficial for guiding and optimizing the analysis of lesions in ultrasound images.

Description

Technical Field

[0001] This invention relates generally to the field of ultrasound imaging technology, and more specifically to an ultrasound imaging system and an ultrasound image analysis method. Background Technology

[0002] The thyroid gland is the largest endocrine gland in the human body, playing a vital role in growth, development, and metabolism. In recent years, the incidence of thyroid nodules has been on the rise. Similarly, the incidence and mortality rates of breast diseases have been steadily increasing, and they have become common diseases threatening women's physical and mental health. Ultrasound examination, due to its non-invasiveness, simplicity, low cost, and repeatability, has become the preferred clinical diagnostic method for breast and thyroid diseases.

[0003] The signs and symptoms of thyroid and breast lesions are complex, and diagnosis is limited by the physician's clinical experience, thus involving a degree of subjectivity. The TI-RADS (Thyroid Imaging Reporting and Data System) assessment criteria, proposed by the American College of Radiology (ACR) in 2017, is currently the most widely used grading and evaluation standard for thyroid ultrasound diagnosis. The TI-RADS assessment criteria standardize the diagnostic reporting of all normal and abnormal imaging findings of the thyroid gland as a whole organ, using unified professional terminology, standardized diagnostic classification, and examination procedures. The BI-RADS (Breast Imaging Reporting and Data System) assessment criteria, proposed by the ACR in 2013, summarizes the ultrasound manifestations of breast lesions.

[0004] The TI-RADS and BI-RADS assessment criteria involve numerous diagnostic rules, which are difficult for junior doctors and doctors in primary care hospitals to operate and memorize. Furthermore, presenting the lesion characteristics determined according to the above criteria in a textual form is not clear or intuitive enough. Summary of the Invention

[0005] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. The summary section of this invention is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0006] To address the shortcomings of existing technologies, the first aspect of this invention provides an ultrasound imaging system, the ultrasound imaging system comprising:

[0007] Ultrasonic probe;

[0008] A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the thyroid region of the subject being tested;

[0009] A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the thyroid region to obtain ultrasound echo signals;

[0010] Processor, used for:

[0011] Acquire ultrasound images by scanning the thyroid region of the subject.

[0012] Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types;

[0013] Determine the TI-RADS score corresponding to the TI-RADS lesion characteristics, and determine the TI-RADS grade of the lesion based on the TI-RADS score;

[0014] A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0015] Based on the determined TI-RADS lesion features and the corresponding TI-RADS scores, a feature graph is generated on the radar image;

[0016] A display for showing the radar chart, the feature graph, and the TI-RADS classification.

[0017] A second aspect of the present invention provides an ultrasound imaging system, the ultrasound imaging system comprising:

[0018] Ultrasonic probe;

[0019] A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the thyroid region of the subject being tested;

[0020] A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the thyroid region to obtain ultrasound echo signals;

[0021] Processor, used for:

[0022] Acquire ultrasound images by scanning the thyroid region of the subject.

[0023] Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types;

[0024] Determine the TI-RADS score corresponding to the TI-RADS lesion features;

[0025] A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0026] Based on the determined TI-RADS lesion features and the corresponding TI-RADS scores, a feature graph is generated on the radar image;

[0027] A display for showing the radar image and the feature graph.

[0028] A third aspect of this invention provides an ultrasound image analysis method, the method comprising:

[0029] Acquire ultrasound images by scanning the thyroid region of the subject.

[0030] Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types;

[0031] Determine the TI-RADS score corresponding to the TI-RADS lesion characteristics, and determine the TI-RADS grade of the lesion based on the TI-RADS score;

[0032] A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0033] Based on the determined TI-RADS lesion features and the corresponding TI-RADS scores, a feature graph is generated on the radar image;

[0034] The radar chart, the feature graph, and the TI-RADS classification are displayed.

[0035] A fourth aspect of this invention provides an ultrasound image analysis method, the method comprising:

[0036] Acquire ultrasound images by scanning the thyroid region of the subject.

[0037] Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types;

[0038] Determine the TI-RADS score corresponding to the TI-RADS lesion features;

[0039] A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0040] Based on the determined TI-RADS lesion features and the corresponding TI-RADS scores, a feature graph is generated on the radar image;

[0041] The radar image and the feature graph are displayed.

[0042] A fifth aspect of this invention provides an ultrasound imaging system, the ultrasound imaging system comprising:

[0043] Ultrasonic probe;

[0044] A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the breast region of the subject being tested;

[0045] A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the breast region to obtain ultrasound echo signals;

[0046] Processor, used for:

[0047] Acquire ultrasound images by scanning the breast region of the subject.

[0048] Detecting lesions in the ultrasound images;

[0049] Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types;

[0050] The BI-RADS classification of the lesion was determined based on a pre-trained BI-RADS classification model;

[0051] A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0052] Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graph is generated on the radar chart; a display is used to show the radar chart, the feature graph, and the BI-RADS rating.

[0053] A sixth aspect of this invention provides an ultrasound image analysis method, the method comprising:

[0054] Acquire ultrasound images by scanning the breast region of the subject.

[0055] Detecting lesions in the ultrasound images;

[0056] Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types;

[0057] The BI-RADS classification of the lesion was determined based on a pre-trained BI-RADS classification model;

[0058] A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0059] Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graph is generated on the radar chart;

[0060] The radar chart, the feature graph, and the BI-RADS classification are displayed.

[0061] A seventh aspect of the present invention provides an ultrasound imaging system, the ultrasound imaging system comprising:

[0062] Ultrasonic probe;

[0063] A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the breast region of the subject being tested;

[0064] A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the breast region to obtain ultrasound echo signals;

[0065] Processor, used for:

[0066] Acquire ultrasound images by scanning the breast region of the subject.

[0067] Detecting lesions in the ultrasound images;

[0068] Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types;

[0069] A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0070] Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graph is generated on the radar chart;

[0071] A display for showing the radar image and the feature graph.

[0072] An eighth aspect of the present invention provides an ultrasound image analysis method, the method comprising:

[0073] Acquire ultrasound images by scanning the breast region of the subject.

[0074] Detecting lesions in the ultrasound images;

[0075] Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types;

[0076] A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0077] Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graph is generated on the radar chart;

[0078] The radar image and the feature graph are displayed.

[0079] The ultrasound image analysis method and ultrasound imaging system according to embodiments of the present invention use radar charts to intuitively present the scoring of lesions, which is beneficial for guiding and optimizing the analysis of lesions in ultrasound images. Attached Figure Description

[0080] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0081] In the attached diagram:

[0082] Figure 1 A schematic block diagram of an ultrasound imaging system according to an embodiment of the present invention is shown;

[0083] Figure 2 This illustration shows a radar chart, feature graph, and TI-RADS classification based on the TI-RADS evaluation standard according to an embodiment of the present invention.

[0084] Figure 3 This illustration shows a radar chart, feature graph, and TI-RADS rating based on the TI-RADS evaluation criteria according to another embodiment of the present invention.

[0085] Figure 4 The diagram illustrates a radar chart, feature graph, and TI-RADS rating drawn in conjunction with the TI-RADS evaluation criteria according to an embodiment of the present invention.

[0086] Figure 5 The radar chart, feature graph, and TI-RADS classification are shown according to an embodiment of the present invention when all TI-RADS scores are 0.

[0087] Figure 6 A display interface according to an embodiment of the present invention is shown;

[0088] Figure 7 A schematic flowchart of an ultrasound image analysis method according to an embodiment of the present invention is shown;

[0089] Figure 8 A schematic flowchart illustrating an ultrasound image analysis method according to another embodiment of the present invention is shown;

[0090] Figure 9 A radar chart, feature graph, and BI-RADS classification based on the BI-RADS evaluation criteria are shown according to an embodiment of the present invention.

[0091] Figure 10A schematic flowchart of an ultrasound image analysis method according to yet another embodiment of the present invention is shown;

[0092] Figure 11 A schematic flowchart of an ultrasound image analysis method according to another embodiment of the present invention is shown. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0094] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0095] It should be understood that the invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0096] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising” and / or “including,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0097] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0098] Below, first refer to Figure 1 An ultrasound imaging system according to an embodiment of this application is described. Figure 1A schematic structural block diagram of an ultrasound imaging system 100 according to an embodiment of this application is shown.

[0099] like Figure 1 As shown, the ultrasound imaging system 100 includes an ultrasound probe 110, a transmitting circuit 112, a receiving circuit 114, a processor 116, and a display 118. The transmitting circuit 112 excites the ultrasound probe 110 to emit ultrasound waves towards the thyroid region of the subject. The receiving circuit 114 controls the ultrasound probe 110 to receive ultrasound echoes returned from the thyroid region to obtain ultrasound echo signals. The processor 116 is used to: acquire ultrasound images obtained by ultrasound scanning the thyroid region of the subject; detect lesions in the ultrasound images and identify TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; determine the TI-RADS score corresponding to the TI-RADS lesion features, and, based on the TI-RADS feature type... The RADS score determines the TI-RADS grade of the lesion; a radar map is generated using at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar map into multiple partitions, each classification axis or each partition represents a TI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing the score; a feature graph is generated on the radar map based on the TI-RADS score corresponding to the determined TI-RADS lesion feature; a display 118 is used to display the radar map, the feature graph, and the TI-RADS grade. Further, the ultrasound imaging system may also include a transmit / receive selection switch 120 and a beamforming circuit 122, and the transmit circuit 112 and the receive circuit 114 can be connected to the ultrasound probe 110 via the transmit / receive selection switch 120.

[0100] The ultrasound imaging system 100 of this invention provides an intelligent auxiliary analysis method for ultrasound examination of the thyroid gland, which can improve the diagnostic efficiency and accuracy of doctors. Specifically, the graphical display of the TI-RADS analysis results of thyroid lesions in the form of a radar chart allows doctors to more intuitively understand the various ultrasound attributes and malignancy of the lesions, providing strong support for doctors to interpret the TI-RADS analysis results of patients in a more concrete way.

[0101] The ultrasonic probe 110 includes multiple transducer elements. These elements can be arranged in a row to form a linear array, or in a two-dimensional matrix to form a planar array. They can also form a convex array. Each transducer element is used to emit ultrasonic waves based on an excitation electrical signal, or to convert received ultrasonic waves into electrical signals. Therefore, each transducer element can be used to achieve the mutual conversion between electrical pulse signals and ultrasonic waves, thereby enabling the emission of ultrasonic waves to the target area of ​​the object being tested, and also to receive ultrasonic wave echoes reflected back from the tissue. During ultrasonic testing, the transmission and reception sequences can be used to control which transducer elements are used to emit ultrasonic waves and which are used to receive ultrasonic waves, or to control the transducer elements to be used in time-slotted manner for emitting ultrasonic waves or receiving ultrasonic wave echoes. Transducer elements participating in ultrasonic wave emission can be simultaneously excited by electrical signals, thus emitting ultrasonic waves simultaneously; alternatively, transducer elements participating in ultrasonic beam emission can be excited by several electrical signals with a certain time interval, thus continuously emitting ultrasonic waves with a certain time interval.

[0102] During ultrasound imaging, the transmitting circuit 112 is used to excite the ultrasound probe 110 to emit ultrasonic waves toward the object under test; the receiving circuit 114 is used to control the ultrasound probe 110 to receive the ultrasonic echo returned from the object under test in order to obtain the ultrasonic echo signal.

[0103] Specifically, during ultrasound imaging, the transmitting circuit 112 sends a delayed-focused transmission pulse to the ultrasound probe 110 via the transmit / receive selection switch 120. Excited by the transmission pulse, the ultrasound probe 110 emits an ultrasonic beam towards the tissue of the target area of ​​the object being measured. After a certain delay, it receives the ultrasonic echo reflecting back from the tissue of the target area, carrying tissue information, and converts this ultrasonic echo back into an electrical signal. The receiving circuit 114 receives the electrical signal converted by the ultrasound probe 110, obtains the ultrasonic echo signal, and sends these ultrasonic echo signals to the beamforming circuit 122. The beamforming circuit 122 performs focusing delay, weighting, and channel summation on the ultrasonic echo data, and then sends it to the processor 116.

[0104] Optionally, the processor 116 can be implemented as software, hardware, firmware, or any combination thereof, and can use one or more application-specific integrated circuits (ASICs), one or more general-purpose integrated circuits, one or more microprocessors, one or more programmable logic devices, or any combination of the foregoing circuits and / or devices, or other suitable circuits or devices. Furthermore, the processor 116 can control other components in the ultrasound imaging system 100 to perform the corresponding steps of the methods in the various embodiments of this specification.

[0105] Processor 116 performs signal detection, signal enhancement, data conversion, and logarithmic compression on the ultrasound echo signal to form an ultrasound image. The ultrasound image obtained by processor 116 can be displayed on display 118 or stored in memory 124. In addition to processing the ultrasound echo signal to generate an ultrasound image of the target area in real time, processor 116 can also acquire ultrasound images of the target area of ​​the subject through other means. For example, processor 116 can retrieve pre-stored ultrasound images of the target area from memory 124, or processor 116 can control the reception of ultrasound images of the target area transmitted from other ultrasound systems or networks. The target area is the body region for ultrasound imaging; for example, in real-time scanning, the target area refers to the body region scanned by the doctor through the probe. In one embodiment, the target area of ​​the subject includes the thyroid region.

[0106] After acquiring ultrasound images, lesions in the ultrasound images can be detected automatically, manually, or semi-automatically.

[0107] When the processor 116 automatically detects lesions in an ultrasound image, it can first mark the approximate location of the lesion region based on a detection algorithm or detection model, and then segment or extract the lesion boundary based on a segmentation algorithm. Detection algorithms include, but are not limited to, algorithms based on deep learning, machine learning, and traditional image processing.

[0108] For example, when using deep learning algorithms, it is necessary to first train the deep learning neural network based on collected sample ultrasound images and the annotation results of lesion regions by senior physicians (i.e., the bounding boxes of the ROI regions, i.e., coordinate information). Deep learning neural networks include, but are not limited to, RCNN, Faster RCNN, SSD, and YOLO. During the network training phase, the error between the lesion detection result and the annotation result is calculated during the iteration process, and the weights in the network are continuously updated with the aim of minimizing the error. This process is repeated continuously, so that the detection result gradually approaches the true value of the lesion region ROI, resulting in a trained ROI detection model. This model can achieve automated lesion detection and extraction for new input data.

[0109] When using a combination of traditional image processing and machine learning for lesion detection, the process begins by identifying candidate regions using image processing methods, such as the Select Search algorithm. Next, the candidate regions are transformed to a fixed size, and image processing techniques are used to extract image features such as gradients and textures, for example, based on SIFT operators, HoG operators, and GLCM (Gray-Level Co-occurrence Matrix). Finally, a trained traditional machine learning algorithm utilizes these image features to obtain the bounding box of the lesion through regression.

[0110] When segmenting or extracting lesion boundaries based on segmentation algorithms, a trained deep learning segmentation model can be used to extract the lesion boundaries from the detected lesion ROI region or directly from the entire ultrasound image. Deep learning segmentation networks include UNet, FCN, and networks improved upon them. For example, when training a deep learning segmentation model, the input samples are an ultrasound image and the labeled region of the lesion in the image. This labeled region can be a binarized image of the lesion, or the lesion location information can be written into a labeling file such as XML or JSON. The error between the segmentation result output by the model and the labeled result is calculated, and the error is iterated until the segmentation result approaches the true value, thus completing the training of the deep learning segmentation model.

[0111] In some embodiments, multi-task deep learning networks that perform simultaneous detection and segmentation can also be used for boundary extraction. Such deep learning networks include mask-RCNN, PolarMask, SOLO, etc. The first step of such deep learning networks is to locate the approximate location of the lesion region, and then to perform fine segmentation of the lesion boundary.

[0112] Traditional image processing algorithms for lesion segmentation can be region-based or gradient-based. Region-based segmentation algorithms include region growing, watershed, and Otsu thresholding; gradient-based segmentation algorithms include the Sobel operator and the Canny operator.

[0113] When using a machine learning-based lesion segmentation method, a machine learning segmentation model can be trained in advance based on the collected ultrasound images and lesion annotation results. Machine learning models such as SVM, Kmeans (K-means clustering algorithm), and Cmeans (C-means clustering algorithm) can be used to perform binary classification on the gray value or texture value of the image pixels to determine whether each pixel or the texture feature vector representing the current pixel belongs to the lesion region, thereby achieving the segmentation or extraction of the lesion region.

[0114] The above illustrates several exemplary methods for automatic lesion detection. In other implementations, the user can manually mark the lesion area in the ultrasound image, for example, by displaying the ultrasound image on monitor 118 and determining the location of the lesion based on the user's manual marking. Alternatively, the location of the lesion can be determined through semi-automatic detection. For instance, the location of the lesion on the ultrasound image can first be automatically detected based on a machine recognition algorithm, and then further modified or corrected by the user to obtain a more accurate location. Any other suitable method can also be used to detect lesions in ultrasound images.

[0115] Subsequently, the processor 116 identifies the lesion features in the ultrasound image corresponding to at least five TI-RADS feature types. Specifically, the TI-RADS assessment criteria proposed by ACR in 2017 include the following five TI-RADS feature types: shape type, composition type, echo type, focal hyperechoic type (also known as hyperechoic type or calcification type), and edge type. Each TI-RADS feature type includes several TI-RADS lesion features, and each TI-RADS lesion feature corresponds to a TI-RADS score. For example, in the TI-RADS assessment criteria proposed by ACR, the lesion features included in each TI-RADS feature type and their TI-RADS scores are as follows:

[0116] a) Shape type: Lesion width greater than height (0 points), lesion height greater than width (3 points);

[0117] b) Composition type: cystic (0 points), spongy (0 points), mixed cystic and solid (1 point), solid (2 points);

[0118] c) Echo type: no echo (0 points), high or equal echo (1 point), low echo (2 points), very low echo (3 points);

[0119] d) Calcification type: no calcification or large comet tail sign (0 points), coarse calcification (1 point), peripheral calcification (2 points), microcalcification (3 points);

[0120] e) Border type: smooth (0 points), unclear outline (0 points), irregular (2 points), invading beyond the thyroid gland (3 points).

[0121] It should be noted that this application does not restrict the version of the TI-RADS evaluation standard. Regardless of which country or organization developed the TI-RADS evaluation standard, whether it is an existing TI-RADS evaluation standard or a future updated TI-RADS evaluation standard, it should be included within the scope of this application. If the future updated TI-RADS evaluation standard includes more TI-RADS feature types, a radar chart can be drawn based on the above five TI-RADS feature types, or a radar chart can be drawn based on the above five TI-RADS feature types and the newly added TI-RADS feature types.

[0122] For example, methods for identifying TI-RADS lesion features under each TI-RADS feature type include, but are not limited to, the following: deep learning-based methods, methods based on traditional image features combined with machine learning, and combinations of the above two methods.

[0123] When identifying TI-RADS lesion features using deep learning methods, a multi-task neural network model can be used to predict TI-RADS lesion features under multiple TI-RADS feature types, or a single-task neural network model can be used to predict TI-RADS lesion features under a single TI-RADS feature type.

[0124] When using a multi-task neural network model to simultaneously predict TI-RADS lesion features under multiple TI-RADS feature types, one embodiment uses the extracted ultrasound image of the lesion region as input. Multiple classification branches of the multi-task deep learning network are directly used to predict the TI-RADS lesion features under each TI-RADS feature type, treating shape type, composition type, echo type, focal hyperechoic type, and edge type as five prediction tasks. For example, when predicting the shape type, the input ultrasound image of the lesion region is processed through shared and private convolutional blocks, and the corresponding classification label is output. The backbone network used in each convolutional block includes, but is not limited to, typical deep learning convolutional classification networks, such as AlexNet, ResNet, and VGG.

[0125] When training such network models, the classification subnet for each TI-RADS feature type can be trained separately, or the entire network can be trained simultaneously. By calculating the error between the prediction results of each branch and the calibration results (the calibration results are the true results of branches such as the edge type, shape type, and echo type of the lesion), the model is continuously iterated and gradually approximated, and finally the classification model and its classification accuracy for each TI-RADS feature type are obtained.

[0126] When using a single-task neural network model to predict TI-RADS lesion features under a single TI-RADS feature type, a deep learning network model can be constructed for each TI-RADS feature type, and multiple deep learning network models can be used in parallel to classify multiple TI-RADS feature types.

[0127] When using a method that combines traditional image features with machine learning for identification, the first step is to use a feature extraction algorithm to extract image features for a single TI-RADS feature type. Then, the extracted image features are classified to obtain the TI-RADS lesion features under that TI-RADS feature type.

[0128] The extracted image features include, but are not limited to, histograms and gray-level co-occurrence matrix features. After extracting the image features, they can be classified based on pre-set thresholds. For example, for echo type, gray-level features can be extracted. When the mean gray level within the lesion is greater than the gray level of the thyroid parenchyma, the lesion's echo type can be classified as hyperechoic or isoechoic. Alternatively, the image features can be concatenated with machine learning models to predict TI-RADS lesion features under the corresponding TI-RADS feature type. For example, for echo type, the extracted image features can be input into machine learning models such as SVM, K-means, and KNN to predict the lesion's echo type and obtain the predicted echo type result.

[0129] When combining the above methods for TI-RADS lesion feature identification, each TI-RADS feature type can be treated as a separate prediction or classification task. For different TI-RADS feature types, an algorithm or model suitable for that specific feature type can be used. This can be either the deep learning-based methods mentioned above, or the methods based on traditional image features combined with machine learning. Alternatively, any other suitable method can be employed for TI-RADS lesion feature identification.

[0130] As described above, each TI-RADS lesion feature in the TI-RADS assessment criteria corresponds to a TI-RADS score. Therefore, after identifying the TI-RADS lesion features under each TI-RADS feature type, the corresponding TI-RADS score can be obtained. In some embodiments, the TI-RADS score of the lesion can also be directly identified. For example, the output of the multi-task neural network model or the single-task neural network model described above can be directly set as the TI-RADS score.

[0131] In addition to determining the TI-RADS score of the lesion, the processor 116 is also used to determine the TI-RADS grade of the lesion based on multiple TI-RADS scores. For example, the processor 116 sums the TI-RADS scores to obtain a total TI-RADS score, and determines the TI-RADS grade of the lesion based on the correspondence between the total TI-RADS score and the TI-RADS grade. Alternatively, the processor 116 may also perform a weighted summation of TI-RADS scores for different TI-RADS feature types, and determine the TI-RADS grade of the lesion based on the result of the weighted summation.

[0132] Specifically, the TI-RADS assessment criteria classify lesions into five TI-RADS grades—TR1, TR2, TR3, TR4, and TR5—based on the sum of their TI-RADS scores. The TR grade represents the degree of malignancy suspicion of the thyroid lesion; a higher TR grade indicates a higher likelihood of malignancy. The mapping relationship between the sum of TI-RADS scores and TI-RADS grades is shown in Table 1.

[0133] Table 1

[0134]

[0135] For example, if the TI-RADS lesion characteristics and TI-RADS scores corresponding to the shape, composition, echo, focal hyperechoicity and edge type of the lesion are identified as solid (2 points), high or isoechoic (1 point), width greater than height (0 points), irregular (2 points) and microcalcification (3 points), then the total TI-RADS score is 8 points. According to the mapping relationship in Table 1, the TI-RADS grade of the lesion can be determined as TR5.

[0136] As described above, the processor 116 determines the TI-RADS score for each TI-RADS feature type of the lesion and the overall TI-RADS grade. Then, the processor 116 draws a radar chart based on at least five TI-RADS feature types and generates a feature graph on the radar chart based on the TI-RADS score corresponding to the determined TI-RADS lesion features. The analysis results obtained from analyzing the lesion based on the TI-RADS evaluation criteria are clearly and intuitively presented through the radar chart and the feature graph displayed on the radar chart.

[0137] A radar chart, also known as a star chart or spider chart, is a two-dimensional graph used to simultaneously display variables in three or more dimensions. Specifically, a radar chart is generated using at least five TI-RADS feature types as classification axes. Each classification axis divides the radar chart into multiple partitions, with each axis or partition representing one of the TI-RADS feature types. At least one axis or partition has a scale unit for representing a score. In some embodiments, exemplarily, the TI-RADS scores on the radar chart gradually increase from the inside out, with 0 points at the very center and 3 points at the edges. By comparing the feature graph with the radar chart, users can quickly understand the TI-RADS score of the lesion under each TI-RADS feature type.

[0138] In some embodiments, the radar chart may further include a base map, with classification axes extending from the center of the base map to its edges, dividing the base map into multiple partitions. In the various forms of radar charts described below, the base map may be circular, polygonal, or other shapes, without limitation. When the base map is polygonal, the number of sides of the polygon may be equal to the number of TI-RADS feature types. That is, since there are five TI-RADS feature types, the base map may be pentagonal, with the five classification axes connecting the center of the pentagon to each vertex. It is understood that, in addition to using the number of sides of the polygon to represent the number of TI-RADS feature types, the number of TI-RADS feature types may also be represented by blocks or partitions corresponding to each side of the polygon, without limitation. When the base map is circular, multiple classification axes may divide the circular base map into five sectors starting from the center.

[0139] In one embodiment, at least one partition has grid lines that divide the partition into at least two sub-intervals to facilitate the determination of the TI-RADS score represented by the feature graph by comparing it to the grid lines. For example, the number of sub-intervals occupied by the feature graph within each partition can represent the corresponding TI-RADS score. Exemplarily, multiple grid lines in each region can be arranged parallel at equal intervals.

[0140] In one embodiment, each partition of the base map represents a TI-RADS feature type. The area of ​​the feature graphic within each partition represents the TI-RADS score for the corresponding TI-RADS feature type. Each partition is identified by the TI-RADS feature type it corresponds to. Users can determine the TI-RADS score for the corresponding TI-RADS feature type based on the area of ​​the feature graphic within each partition; a larger area indicates a higher TI-RADS score.

[0141] See Figure 2 The diagram shows a radar chart where each partition represents a TI-RADS feature type. Figure 2 The radar chart shown has a circular background, but in other implementations, the background shape can be replaced with a pentagon, or the background can be omitted. Figure 2In the radar chart shown, five classification axes 230 divide the base map 210 into five partitions 220. Each partition 220 corresponds to a TI-RADS feature type, and the corresponding TI-RADS feature type is labeled outside each partition 220. Partition 220 displays feature graphs 240 representing the TI-RADS score for the corresponding TI-RADS feature type; the larger the area of ​​the feature graph 240, the higher the corresponding TI-RADS score. The classification axes between partitions representing component type and shape type are marked with scale units for representing the scores. These scale units can be used to determine the score represented by the feature graph in each partition; the scale unit can also be marked at the center of each partition.

[0142] Among them, the TI-RADS scores corresponding to shape, composition, echo, calcification and edge type are 0, 2, 1, 3 and 2 respectively. The area of ​​the feature graphic of each partition depends on the TI-RADS score of each region. That is, the feature graphic area of ​​the partition corresponding to the calcification type is the largest, and the feature graphic area of ​​the partition corresponding to the shape type is 0, that is, no feature graphic is displayed.

[0143] In one embodiment, feature graphics within different partitions can be displayed as different colors or patterns to facilitate differentiation between different feature graphics. For example, in Figure 2 In the radar image shown, the feature patterns corresponding to calcification, echo, edge, and composition type can be displayed in green, yellow, red, and blue, respectively; of course, the feature patterns can also use any other suitable colors or patterns.

[0144] In another embodiment, each category axis of the radar chart is used to represent a TI-RADS feature type, and the feature graph is a graph formed by connecting the coordinate points of the TI-RADS score representing the TI-RADS feature type corresponding to each category axis, with each category axis marked with the TI-RADS feature type corresponding to that category axis.

[0145] See Figure 3 The diagram shows a radar chart for each partition corresponding to a TI-RADS feature type. Figure 3 The radar chart shown has a pentagonal base image, but in other implementations, the base image shape can be replaced with a circle, or it can be omitted entirely. Figure 3In the radar chart shown, five classification axes 320 divide the base map 310 into five partitions 330. Each classification axis 320 corresponds to a TI-RADS feature type, and the corresponding TI-RADS feature type is labeled at the vertices of the classification axes 320. A feature graph 340 is formed by connecting the coordinate points representing the actual TI-RADS score of the lesion on each classification axis 320. The TI-RADS score for each TI-RADS feature type can be determined by referring to the shape of the feature graph 340. For example, if the feature graph protrudes at the classification axis representing the shape, it can be determined that the lesion's shape indicates the highest TI-RADS score for that shape type.

[0146] In one embodiment, the base map and classification axis of the radar chart can be plotted in conjunction with the TI-RADS evaluation criteria, so that the radar chart provides more information about the TI-RADS evaluation criteria.

[0147] For example, in some embodiments, the first classification axis in the classification axis includes a maximum scale unit, a minimum scale unit, and a preset scale unit, wherein the preset scale unit is used to represent the maximum TI-RADS score that can be obtained for each TI-RADS lesion feature under the first TI-RADS feature type corresponding to the first classification axis; the above-mentioned ultrasound image analysis method further includes: when the preset scale unit is smaller than the maximum scale unit and larger than the minimum scale unit, distinguishing the portion between the maximum scale unit and the preset scale unit in the first classification axis from the portion between the minimum scale unit and the preset scale unit. This distinguishing display may include: drawing the portion between the minimum scale unit and the preset scale unit as a solid line, and drawing the portion between the maximum scale unit and the preset scale unit as a dashed line. For example, see... Figure 4 For a given component type, the TI-RADS scores for lesion features under that type are only 0, 1, and 2; there are no TI-RADS lesion features with a score of 3. For example, the first classification axis can be a coordinate axis corresponding to that component type. The highest TI-RADS score for that component type is 2, which is less than the maximum scale unit of 3 points corresponding to the first classification axis. Therefore, the portion between 2 and 3 points on the first classification axis can be displayed as a dashed line to indicate that the highest TI-RADS score for that component type is 2.

[0148] In some embodiments, since not all TI-RADS feature types contain all TI-RADS lesion features with all scores, for example, the shape type only has lesion features with scores of 0 and 3, and the composition type only has lesion features with scores of 0, 1, and 2; therefore, the scores of the TI-RADS lesion features existing under each TI-RADS feature type can be represented by grid lines; in this case, the score scale of the TI-RADS score can be marked on the classification axis corresponding to the TI-RADS feature type containing all scores. Specifically, when TI-RADS lesion features with the same TI-RADS score exist simultaneously under TI-RADS feature types corresponding to adjacent classification axes, grid lines can be used to connect the coordinate points representing the same TI-RADS score. For example, if TI-RADS lesion features with a score of 1 exist simultaneously under TI-RADS feature types corresponding to adjacent classification axes, then grid lines are used to connect the coordinate points with a score of 1 on adjacent coordinate axes. When TI-RADS lesion features with the same TI-RADS score do not exist simultaneously under TI-RADS feature types corresponding to adjacent classification axes, then no grid lines are set. For example, if a TI-RADS lesion feature with a score of 1 exists under a TI-RADS feature type corresponding to a classification axis, but a TI-RADS lesion feature with a score of 1 does not exist under the TI-RADS feature type corresponding to an adjacent classification axis, then no grid line connects the two coordinate points.

[0149] Continue to refer to Figure 4 In the marginal type, TI-RADS scores of 0, 2, and 3 exist, but there are no TI-RADS lesion features with a score of 1. Therefore, the classification axis corresponding to the marginal type only marks coordinate points representing scores of 0, 2, and 3, and does not mark coordinate points representing scores of 1. In the strong echo type, all TI-RADS scores exist, and therefore, the classification axis corresponding to the strong echo type is marked with coordinate points representing scores of 0, 1, 2, and 3. Since there are no coordinate points representing scores of 1 on the classification axis corresponding to the marginal type, a grid line connects the classification axes corresponding to the marginal type and the strong echo type between the coordinate points representing scores of 2, but there is no grid line connecting the coordinate points representing scores of 1.

[0150] In one embodiment, when using a radar chart where each classification axis corresponds to one TI-RADS feature type as described above, the origin of the coordinates representing a TI-RADS score of 0 on the classification axis can be offset from the center of the radar chart, i.e., from the intersection of the extended lines of all classification axes. Therefore, when all TI-RADS scores are 0, the feature graph is a small graphic located at the center of the base map, formed by connecting the origins of the coordinates on each classification axis. This avoids the feature graph completely disappearing when all TI-RADS scores are 0, thus preventing a negative impact on the user experience. Figure 5 The radar chart shows the situation when all TI-RADS scores are 0. The small pentagon at the center of the radar chart represents the feature graphic at this time. The feature graphic at this time can be distinguished from the original small pentagon in the background chart, for example, by changing the color of the small pentagon to indicate that the small pentagon is the feature graphic at this time.

[0151] In one embodiment, portions of the feature graph corresponding to different TI-RADS scores can also be displayed in different colors. For example, the color of the feature graph can be set to a gradient from light to dark from the center to the edge, a gradient from dark to light, or a gradient from one color to another. Users can determine the degree to which the lesion is malignant according to the TI-RADS lesion features under each TI-RADS feature type based on the degree of color gradient in the feature graph. When using... Figure 2 When each region corresponds to a radar map of a TI-RADS feature type, the color of the edge of the feature pattern within each region reflects the malignancy of the TI-RADS lesion feature; when using, as shown in the radar map... Figure 3 When each classification axis corresponds to a TI-RADS feature type in the radar chart, the color near the intersection of the feature graph and each classification axis reflects the malignancy of the TI-RADS lesion feature.

[0152] For example, targeting Figure 3 The radar chart shown illustrates a feature image where the color of the feature graphic changes from light to dark from the inside out. Assuming the feature graphic is green, since the TI-RADS score for the shape type is 3, the color of the feature graphic changes from light green near the coordinate point representing 0 points to dark green near the coordinate point representing 3 points along the classification axis corresponding to the shape type. Conversely, the TI-RADS score for the edge type is 0, so the feature graphic is light green near the intersection with the classification axis corresponding to the edge type.

[0153] In one embodiment, the color of the background image, the color of the feature graph, the color of the classification axis, the color of the grid lines, or the color of the border of the TI-RADS grading result can be determined based on the TI-RADS grading of the lesion or the sum of the TI-RADS scores. For example, if the TI-RADS grading result is TR5, highly suggestive of malignancy, the background image can be displayed as dark red; if the TI-RADS grading result is TR3, suggesting benignity, the background image can be displayed as a soft light blue. The color of the background image can refer to the background color of the background image or the color of the background image edges. The color of the feature graph can refer to the background color of the feature graph or the color of the feature graph edges.

[0154] In the radar image of this embodiment, since the area of ​​the feature graphic is affected by various TI-RADS scores, the ratio of the area of ​​the feature graphic to the area of ​​the base image can be calculated and displayed, reflecting the overall score of the lesion. The sum of the TI-RADS scores is a scalar value of 0-15 points, directly proportional to the malignancy of the lesion. However, many doctors may not be aware of the correspondence between the TI-RADS score and the malignancy of the lesion. The aforementioned ratio is a probability value between 0-100% obtained by discretizing the TI-RADS score through the radar image. The higher the malignancy of the lesion, the closer the ratio is to 100%. It is more intuitive than the 0-15 TR total score and easier for doctors to understand. For example, the ratio can be displayed around the radar image, or it can be displayed around the lesion in the ultrasound image.

[0155] In one embodiment, the radar image and feature graph can be displayed simultaneously with at least one of the following: ultrasound image, thyroid position diagram, region of interest in the ultrasound image, boundary of lesion detected in the ultrasound image, and personal information of the subject, to facilitate user comparison, browsing, and comprehensive analysis. Some information can be displayed on the main screen, and some on the touchscreen. Figure 6 An exemplary display interface is shown that simultaneously displays a radar map and feature graph 610, a thyroid positional image 620, an ultrasound image 630, and the boundary 640 of a lesion detected in the ultrasound image. When a user operates on the radar map and feature graph 610, the thyroid positional image 620 and the ultrasound image 630 can display effects corresponding to the user's operations.

[0156] For example, when a user selects a location in the radar chart or feature graph corresponding to each TI-RADS feature type, the ultrasound image can be displayed with an effect corresponding to the selected TI-RADS feature type, allowing the user to view the ultrasound image according to the TI-RADS. The location in the radar chart or feature graph corresponding to each TI-RADS feature type may include a partition corresponding to each TI-RADS feature type, a feature graph corresponding to each TI-RADS feature type, text identifying each TI-RADS feature type, etc.

[0157] For example, when the user selects a location in the radar image corresponding to a shape type or edge type, the boundary of the lesion area in the ultrasound image can be highlighted, allowing the user to view the actual shape or edge of the lesion according to the TI-RADS score for the shape type or edge type. When the user selects a location in the radar image corresponding to an echo type or composition type, the entire lesion area in the ultrasound image can be highlighted. When the user selects a location in the radar image corresponding to a calcification type, the calcified area extracted from the ultrasound image can be highlighted. This highlighting can be a flashing display or a high-brightness display, etc.

[0158] Display 118 is used to display the radar chart, the feature graph, and the TI-RADS grading. The radar chart and feature graph can present the TI-RADS score for each TI-RADS feature type, and the TI-RADS grading indicates the overall benign or malignant degree of the lesion. For example, Figure 2 , Figure 3 The lesion shown in the radar image has a TI-RADS grade of TR5. Figure 4 The lesion shown in the radar image is classified as TR4 according to the TI-RADS standard. Figure 5 The lesion shown around the radar image is classified as TR0 according to the TI-RADS classification.

[0159] The display 118 is connected to the processor 116. The display 118 can be a touch screen, an LCD screen, etc.; or, the display 118 can be an independent display such as an LCD screen or a television, separate from the ultrasound imaging system 100; or, the display 118 can be the screen of an electronic device such as a smartphone or tablet, etc. The number of displays 118 can be one or more. For example, the display 118 may include a main screen and a touch screen, with the main screen primarily used to display ultrasound images and the touch screen primarily used for human-computer interaction.

[0160] The display 118, while displaying the radar chart, the feature graph, and the TI-RADS classification, can also provide a graphical interface for human-computer interaction. One or more controlled objects can be set on the graphical interface, allowing the user to input operation commands via a human-computer interaction device to control these controlled objects and execute corresponding control operations. For example, icons can be displayed on the graphical interface, and the human-computer interaction device can be used to operate these icons to perform specific functions, such as drawing a region of interest bounding box on an ultrasound image.

[0161] Optionally, the ultrasound imaging system 100 may also include other human-machine interface devices besides the display 118, which are connected to the processor 116. For example, the processor 116 may be connected to the human-machine interface device via an external input / output port, which may be a wireless communication module, a wired communication module, or a combination of both. The external input / output port may also be based on USB, bus protocols such as CAN, and / or wired network protocols.

[0162] The human-computer interaction device may include an input device for detecting user input information. This input information may be, for example, control commands for the timing of ultrasonic wave transmission / reception, operational input commands for drawing points, lines, or boxes on an ultrasonic image, or other types of commands. The input device may include one or a combination of several of the following: a keyboard, mouse, scroll wheel, trackball, mobile input device (such as a mobile device with a touchscreen, a mobile phone, etc.), a multi-function knob, etc. The human-computer interaction device may also include an output device such as a printer.

[0163] The ultrasound imaging system 100 may also include a memory 124 for storing instructions executed by the processor, storing received ultrasound echoes, storing ultrasound images, etc. The memory may be a flash memory card, solid-state memory, hard disk, etc. It may be volatile and / or non-volatile memory, removable memory and / or non-removable memory, etc.

[0164] It should be understood that Figure 1 The components included in the ultrasound imaging system 100 shown are merely illustrative and may include more or fewer components, which is not limited in this application.

[0165] Below, we will refer to Figure 7 An ultrasound image analysis method according to an embodiment of the present invention is described. Figure 7 This is a schematic flowchart of an ultrasound image analysis method 700 according to an embodiment of the present invention. Figure 7 As shown, the ultrasound image analysis method 700 of this embodiment includes the following steps:

[0166] In step S710, an ultrasound image is obtained by performing an ultrasound scan on the thyroid region of the subject.

[0167] In step S720, lesions in the ultrasound image are detected, and TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types are identified;

[0168] In step S730, the TI-RADS score corresponding to the TI-RADS lesion feature is determined, and the TI-RADS grade of the lesion is determined based on the TI-RADS score;

[0169] In step S740, a radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein, the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0170] In step S750, a feature graph is generated on the radar image based on the TI-RADS score corresponding to the determined TI-RADS lesion features;

[0171] In step S760, the radar chart and the TI-RADS classification are displayed. The ultrasound image analysis method 200 of this embodiment can be implemented by the ultrasound imaging system 100 described above. The relevant descriptions of each step can be found in the above description of the ultrasound imaging system 100, and will not be repeated here.

[0172] The ultrasound imaging system and ultrasound image analysis method 700 according to embodiments of the present invention use radar charts to intuitively present the TI-RADS scores of lesions, and simultaneously display the TI-RADS classification of lesions, which is beneficial for guiding and optimizing the analysis of lesions in ultrasound images.

[0173] The following continues to refer to... Figure 1 The present application describes an ultrasound imaging system according to another embodiment of the present application. The ultrasound imaging system includes an ultrasound probe 110, a transmitting circuit 112, a receiving circuit 114, a processor 116, and a display 118. The relevant descriptions of each component can be referred to the relevant description of the ultrasound imaging system 100 above. The following only describes the main functions of the ultrasound imaging system, and the details already described above are omitted.

[0174] Specifically, the transmitting circuit 112 is used to excite the ultrasound probe 110 to emit ultrasound waves toward the thyroid region of the subject; the receiving circuit 114 is used to control the ultrasound probe 110 to receive the ultrasound echo returned from the thyroid region to obtain an ultrasound echo signal; the processor 116 is used to: acquire an ultrasound image obtained by performing an ultrasound scan on the thyroid region of the subject; detect lesions in the ultrasound image and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; determine the TI-RADS score corresponding to the TI-RADS lesion features; generate a radar chart using at least five TI-RADS feature types as classification axes; wherein the classification axis divides the radar chart into multiple partitions, each classification axis or each partition is used to represent a TI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing the score; generate a feature graph on the radar chart based on the TI-RADS score corresponding to the determined TI-RADS lesion features; and the display 118 is used to display the radar chart and the feature graph.

[0175] The ultrasound imaging system in this embodiment is largely similar to the ultrasound imaging system described above; some identical details are omitted below. This ultrasound imaging system also generates a radar map based on TI-RADS feature types and generates a feature graph on the radar map based on the TI-RADS score of the lesion under each TI-RADS feature type. The specific forms of the radar map and feature graph can be found above. Unlike the ultrasound imaging systems described above, this ultrasound imaging system does not limit the determination of the TI-RADS grade of the lesion based on the sum of the TI-RADS scores; therefore, it does not limit the correlation between the TI-RADS grade and the radar map. Figure 1 Similarly, it is only necessary to present the TI-RADS score of the lesion under each TI-RADS feature type through a radar chart.

[0176] In one embodiment, in addition to the radar chart, the sum of all TI-RADS scores may be displayed. A higher sum of TI-RADS scores indicates a greater likelihood that the lesion is malignant. In another embodiment, the radar chart also includes a base map, with the classification axis extending from the center region of the base map to its edges to divide the base map into multiple partitions. The method further includes calculating and displaying the ratio of the area of ​​the feature pattern in the radar chart to the area of ​​the base map. A higher ratio indicates a greater likelihood that the lesion is malignant.

[0177] Figure 8 A schematic flowchart of an ultrasound image analysis method 800 according to another embodiment of this application is shown. Figure 8 As shown, the ultrasound image analysis method 800 includes the following steps:

[0178] Step S810: Obtain an ultrasound image obtained by performing an ultrasound scan on the thyroid region of the subject.

[0179] Step S820: Detect lesions in the ultrasound image and identify TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types;

[0180] Step S830: Determine the TI-RADS score corresponding to the TI-RADS lesion features;

[0181] Step S840: Generate a radar chart using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score.

[0182] Step S850: Based on the TI-RADS score corresponding to the determined TI-RADS lesion features, a feature graph is generated on the radar image;

[0183] Step S860: Display the radar image and the feature pattern.

[0184] The ultrasound image analysis method 800 of this invention can be implemented by the ultrasound imaging system described above. The relevant descriptions of each step can be referred to the relevant descriptions above, and will not be repeated here.

[0185] The ultrasound image analysis method 800 and ultrasound imaging system according to embodiments of this application use radar charts to intuitively present the scoring of lesions, which is beneficial for guiding and optimizing the analysis of lesions in ultrasound images.

[0186] Another aspect of this application provides an ultrasound imaging system. Continuing to refer to... Figure 1 The ultrasound imaging system includes an ultrasound probe 110, a transmitting circuit 112, a receiving circuit 114, a processor 116, and a display 118. The relevant descriptions of each component can be found in the description of the ultrasound imaging system 100 above. The following only describes the main functions of the ultrasound imaging system and omits the details already described above.

[0187] Specifically, the transmitting circuit 112 is used to excite the ultrasound probe 110 to emit ultrasound waves toward the breast region of the subject; the receiving circuit 114 is used to control the ultrasound probe 110 to receive the ultrasound echo returned from the curve region to obtain an ultrasound echo signal; the processor 116 is used to: acquire an ultrasound image obtained by performing an ultrasound scan on the breast region of the subject; detect lesions in the ultrasound image; determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; determine the BI-RADS grade of the lesion based on a pre-trained BI-RADS grading model; generate a radar chart using the at least seven BI-RADS feature types as the classification axis of the radar chart; wherein the classification axis divides the radar chart into multiple partitions, each classification axis or each partition is used to represent a BI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing the score; generate a feature graph on the radar chart based on the BI-RADS score corresponding to the determined BI-RADS feature type; and the display 118 is used to display the radar chart, the feature graph, and the BI-RADS grade.

[0188] The ultrasound imaging system of this invention provides an intelligent auxiliary analysis method for breast ultrasound examination, which can improve the diagnostic efficiency and accuracy of doctors. Specifically, the graphical display of BI-RADS analysis results of breast lesions in the form of radar charts allows doctors to more intuitively understand the various ultrasound attributes and malignancy levels of the lesions, providing strong support for doctors to interpret the patient's BI-RADS analysis results more concretely.

[0189] Specifically, the ultrasound image obtained by the processor 116 can be obtained by real-time ultrasound scanning of the breast region of the subject, or by retrieving a pre-stored ultrasound image of the breast region of the subject from the memory; or it can receive ultrasound images of the breast region of the subject transmitted from other ultrasound systems or networks.

[0190] After obtaining ultrasound images, lesions in the ultrasound images can be detected automatically, manually, or semi-automatically. The methods for detecting lesions are detailed above regarding the detection methods for thyroid lesions and will not be repeated here.

[0191] Subsequently, processor 116 determines the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types. The BI-RADS assessment criteria proposed by ACR include the following seven BI-RADS feature types: orientation type, shape type, edge type, internal echo type, posterior echo type, calcification type, and blood flow type. Each BI-RADS feature type includes several BI-RADS lesion features. Specifically, the BI-RADS lesion features included in each BI-RADS feature type are as follows:

[0192] Direction type: parallel, non-parallel;

[0193] Shape types: oval, circle, irregular shape;

[0194] Edge types: sharp, blurry, angular, slightly lobed, burrs;

[0195] Internal echo types: anechoic, isoechoic, cystic-solid mixed echo, hypoechoic, heterogeneous echo, hyperechoic;

[0196] Posterior echo type: enhanced, unchanged, attenuated, mixed change;

[0197] Calcification type: No calcification, calcification present;

[0198] Blood flow type: no blood flow, peripheral blood flow, internal blood flow.

[0199] It should be noted that this application does not limit the version of the BI-RADS evaluation standard. Regardless of which country or organization developed the BI-RADS evaluation standard, whether it is an existing BI-RADS evaluation standard or a future updated BI-RADS evaluation standard, it should be included within the scope of this application. If the future updated BI-RADS evaluation standard includes more BI-RADS feature types, a radar chart can be drawn based on the above seven BI-RADS feature types, or a radar chart can be drawn based on the above seven BI-RADS feature types and the newly added BI-RADS feature types.

[0200] Since the current BI-RADS assessment criteria do not clearly specify the specific contribution of different BI-RADS lesion features to the BI-RADS grading of breast lesions, but only indicate that different BI-RADS lesion features indicate different degrees of malignancy, the embodiments of this application first perform quantitative analysis on each BI-RADS feature type to obtain the corresponding BI-RADS score.

[0201] In one embodiment, the BI-RADS score for each BI-RADS feature type of the lesion is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features under each BI-RADS feature type. The following description mainly takes the preset lesion state as a malignant lesion state as an example, but the preset lesion state can also be a benign lesion state or other lesion states.

[0202] Specifically, the seven BI-RADS feature types—direction, shape, edge, internal echo, posterior echo, calcification, and blood flow—can be considered as seven classification problems. For each BI-RADS feature type, various feature construction algorithms can be used to extract relevant image features. Based on each type of image feature (e.g., image features related to internal echo), the probability of the lesion being in a preset lesion state is predicted separately, and this probability is used as the BI-RADS score for the current BI-RADS feature type. Seven BI-RADS scores can be obtained for the seven BI-RADS feature types.

[0203] In one embodiment, quantitative evaluation of BI-RADS scores can be achieved based on a multi-task mechanism of a deep learning network. Specifically, the ultrasound image is input into a pre-trained deep learning model for each BI-RADS feature type. The BI-RADS feature prediction branch of the deep learning model is used as the backbone network, from which probability prediction branches are differentiated. The probability prediction branch of the deep learning model predicts the probability for each BI-RADS feature type, and the feature prediction branch of the deep learning model predicts the BI-RADS lesion features for each BI-RADS feature type.

[0204] The backbone network includes, but is not limited to, typical feature extraction networks such as AlexNet, VGG, and ResNet. Taking internal echo type as an example, the backbone branch of the model is used to predict the internal echo category. The feature map of the output layer at the end of the network is the feature map generated when this branch classifies different internal echo types. This feature map is extracted and separately processed through several private convolutional blocks to predict whether the lesion is benign or malignant, obtaining the probability that the lesion is predicted as a malignant lesion based on the image features related to internal echoes. When training the model, the internal echo type prediction branch can be trained separately to achieve a certain accuracy first, and then the entire network can be fine-tuned.

[0205] In another embodiment, image features of the ultrasound image can be extracted for each BI-RADS feature type based on a feature extraction algorithm. Based on the image features, a pre-trained probability prediction machine learning model is used to predict the probability under each BI-RADS feature type, and a pre-trained feature prediction machine learning model is used to predict the BI-RADS lesion features under each BI-RADS feature type.

[0206] Taking internal echo type as an example, firstly, image features related to internal echo, such as lesion gray level, gray level decay rate, and entropy, are extracted. Then, machine learning algorithms such as SVM, random forest, and logistic regression are used to predict the probability of classifying the lesion as a malignant lesion based on the current image features. At the same time, based on these features and other prediction models, BI-RADS feature types are classified, such as dividing internal echo type into specific subclasses such as hyperechoic and isoechoic.

[0207] In addition, a combined approach can be used. Specifically, deep learning models and feature extraction algorithms are used to extract image features from the ultrasound image separately. The image features extracted by the deep learning model and the feature extraction algorithm are then fused to obtain fused image features corresponding to each BI-RADS feature type. Probabilities are then predicted based on these fused image features. Feature fusion methods are not limited to operations such as stitching or dimensionality reduction.

[0208] In another embodiment, the BI-RADS score for each BI-RADS feature type of the lesion represents the contribution of the BI-RADS lesion feature under each BI-RADS feature type to the BI-RADS grade of the lesion. Specifically, the BI-RADS assessment criteria classify breast lesions into seven grades (1, 2, 3, 4a, 4b, 4c, and 5) according to their malignancy. When the BI-RADS score represents the contribution of the BI-RADS lesion feature under each BI-RADS feature type to the BI-RADS grade of the lesion, step S940 is executed simultaneously with step S930, i.e., the BI-RADS grade of the lesion is determined.

[0209] For example, firstly, image features of the ultrasound image under each BI-RADS feature type are extracted, and then the image features under each BI-RADS feature type are integrated into a high-dimensional feature. Image feature extraction can employ traditional image processing methods, machine learning, or deep learning methods, and multiple image features can be extracted for each BI-RADS feature type.

[0210] Next, the BI-RADS grading of the lesions is predicted based on this high-dimensional feature. For example, a random forest classifier can be used to predict the BI-RADS grading based on this high-dimensional feature. Then, based on the image features under each BI-RADS feature type, a machine learning model is used to evaluate the contribution of each BI-RADS lesion feature under that BI-RADS feature type to the BI-RADS grading. Specifically, in one example, the Gini index of the image features under each BI-RADS feature type can be obtained based on the random forest model; the GINI index can evaluate the importance score of each image feature under each BI-RADS feature type in the random forest classifier, reflecting the correlation between different image features and the BI-RADS grading results obtained by the classifier. For example, after obtaining the GINI index for each image, the scores of all features are divided by the total number of features to perform a normalization operation. Finally, the GINI index corresponding to each image feature related to each BI-RADS feature type is weighted separately to obtain the final BI-RADS score for each BI-RADS feature type, which is used to reflect the contribution of the BI-RADS lesion features of the corresponding BI-RADS feature type to the BI-RADS classification.

[0211] The above illustrates an example of calculating the contribution of BI-RADS lesion features to BI-RADS grading based on the GINI index. In another embodiment of this application for calculating the contribution of BI-RADS lesion features to BI-RADS grading, image features of the ultrasound image under each BI-RADS feature type can be extracted using various feature extraction methods. A logistic regression prediction model is then used to fit the contribution value of BI-RADS lesion features to BI-RADS grading for each BI-RADS feature type. The image feature extraction methods can be understood by referring to the aforementioned related descriptions, and will not be repeated here.

[0212] The decision function of this logistic regression prediction model can be either a linear or a nonlinear decision function. The decision function can be represented by Equation 1 below:

[0213]

[0214] Where z is the decision function, as shown in formula (1), the decision function z can be mapped to a probability value g(z) in the interval [0,1] through the sigmoid function, and g(z) represents the BI-RADS classification result.

[0215] In embodiments of the present invention, the decision function z includes, but is not limited to, a linear decision function, as shown in Formula 2 below:

[0216] z(x) = w0x0 + w1x1 + ... + w n x n Formula 2

[0217] Where, [x1,x2,x3,…,x n ] represents the input breast ultrasound image features; x n Represents eigenvalues; w n For different eigenvalues ​​x n The corresponding regression coefficients, i.e., the eigenvalues ​​x n The contribution weights for the current BI-RADS hierarchical prediction problem. In the embodiments of this application, the contribution coefficient w can be... n Perform standardization and normalization operations.

[0218] Finally, the contribution coefficients of each image feature related to each BI-RADS feature type are weighted separately to obtain the contribution value of BI-RADS lesion features to BI-RADS classification under each BI-RADS feature type.

[0219] The processor 116 is also used to determine the BI-RADS grade of the lesion based on a pre-trained BI-RADS grading model. As described above, if the BI-RADS score is the contribution of the BI-RADS lesion feature under each BI-RADS feature type to the BI-RADS grade of the lesion, then the BI-RADS grade of the lesion is determined in the process of determining the contribution. If the BI-RADS score is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion feature under each BI-RADS feature type, then other pre-trained BI-RADS grading models can be used to determine the BI-RADS grade of the lesion.

[0220] Next, processor 116 draws a radar chart based on BI-RADS feature types and generates a feature graph on the radar chart based on the BI-RADS scores corresponding to the BI-RADS feature types. The form of the radar chart and feature graph is similar to the radar chart used to present BI-RADS scores mentioned above, the main difference being the different quantitative indicators used to draw the radar chart and feature graph. Specifically, the radar chart includes a base map, multiple classification axes dividing the base map into multiple partitions, and a feature graph representing the BI-RADS score for each BI-RADS feature type. Specifically, the radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein, the classification axis divides the radar chart into multiple partitions, each classification axis or each partition represents a BI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing the score.

[0221] In one embodiment, the radar chart further includes a base map, with the classification axis extending from the central region of the base map to its edge to divide the base map into multiple partitions. The base map can be circular or polygonal. When the base map is polygonal, the number of sides of the polygon equals the number of BI-RADS feature types. Since the evaluation criteria proposed by ACR include seven BI-RADS feature types, the base map can be heptagonal.

[0222] For example, at least one partition has grid lines that divide the partition into at least two sub-partitions. When the base map is circular, the grid lines are arcs; when the base map is polygonal, the grid lines are straight lines parallel to each side of the polygon.

[0223] In one embodiment, each partition of the base map is used to represent a BI-RADS feature type, the area of ​​the feature graphic within each partition represents the BI-RADS score of the BI-RADS feature type corresponding to that partition, and each partition is identified by the BI-RADS feature type corresponding to that partition.

[0224] See Figure 9 This shows a radar chart where each partition of the base map represents a BI-RADS feature type. Figure 9 In the radar chart shown, seven classification axes 920 divide the base map 910 into seven partitions 930. The area of ​​the feature pattern 940 within each partition 930 represents the BI-RADS score for the corresponding BI-RADS feature type. This BI-RADS score indicates the probability that the lesion is malignant according to the BI-RADS lesion feature under the corresponding BI-RADS feature type. Figure 9 In the example, each partition 930 is labeled with the corresponding BI-RADS feature type, and the classification axis between blood flow type and shape type is labeled with the scale unit of BI-RADS score.

[0225] exist Figure 9 In the radar chart shown, the feature patterns in different zones can be displayed in different colors or patterns to facilitate the differentiation of feature patterns in different zones and to help determine the area of ​​feature patterns in each zone.

[0226] In another embodiment, each category axis of the radar chart is used to represent a BI-RADS feature type, and the feature graph is a graph formed by connecting the coordinate points on each category axis that represent the BI-RADS feature type corresponding to that category axis, with each category axis marked with the BI-RADS feature type corresponding to that category axis.

[0227] In one embodiment, portions of the feature map corresponding to different BI-RADS scores can be displayed in different colors. In another embodiment, the color of the feature map can be determined based on the BI-RADS grading of the lesion. Additionally, the color of the background map can be determined based on the BI-RADS grading of the lesion.

[0228] In one embodiment, when the BI-RADS score for each BI-RADS feature type of the lesion is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion feature under each BI-RADS feature type, since the greater the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion feature under each BI-RADS feature type, the larger the area of ​​the feature graphic, the ratio of the area of ​​the feature graphic to the area of ​​the base map can also be calculated and displayed. This ratio can reflect the probability that the lesion is in the preset lesion state, for example, the probability that the lesion is a malignant lesion.

[0229] In some embodiments, the radar image or feature graph may be displayed simultaneously with at least one of the following: the ultrasound image, a breast position map, a region of interest in the ultrasound image, the boundary of a lesion detected in the ultrasound image, and personal information of the subject.

[0230] When the radar chart or feature graphic is displayed simultaneously with the ultrasound image, the processor 116 is further configured to: when it is determined that the user has selected a position in the radar chart or feature graphic corresponding to each BI-RADS feature type, control the display 118 to display an effect in the ultrasound image that corresponds to the BI-RADS feature type. The position in the radar chart or feature graphic corresponding to each BI-RADS feature type may include a partition corresponding to each BI-RADS feature type, a feature graphic corresponding to each BI-RADS feature type, text identifying each BI-RADS feature type, etc.

[0231] For example, when processor 116 determines that the user has selected a location in a radar map or feature pattern corresponding to a shape type or edge type, it can control display 118 to highlight the boundary of the lesion region in the ultrasound image to present the shape or boundary of the lesion region in the ultrasound image. When processor 116 determines that the user has selected a location in a radar map or feature pattern corresponding to an edge type, internal echo type, or posterior echo type, it can control display 118 to highlight the lesion region in the ultrasound image, for example, by flashing the entire lesion region. When processor 116 determines that the user has selected a location in a radar map or feature pattern corresponding to a calcification type, it can control display 118 to highlight the calcified region extracted in the ultrasound image, for example, by flashing the calcified region in the ultrasound image. When processor 116 determines that the user has selected a location in a radar map or feature pattern corresponding to a blood flow type, it can extract a color blood flow image corresponding to the ultrasound image and control display 118 to overlay the color blood flow image on the ultrasound image. This color blood flow image can be obtained by performing blood flow imaging on the region of interest in the ultrasound image after acquiring the ultrasound image.

[0232] Display 118 is used to display the radar chart, the feature graph, and the BI-RADS classification. See, for example, [link to relevant documentation]. Figure 9 It can display the BI-RADS grade of lesions around the radar image. Figure 9 The lesion shown is classified as BI-RADS 4b. The radar chart and feature graphs present the quantitative results of the BI-RADS score for each BI-RADS feature type. The BI-RADS grading indicates the overall benign or malignant nature of the lesion. Users can perform a combined analysis of both to obtain more accurate and comprehensive results.

[0233] Below, we will refer to Figure 10 An ultrasound image analysis method according to another aspect of an embodiment of the present invention is described. Figure 10 This is a schematic flowchart of an ultrasound image analysis method 1000 according to an embodiment of the present invention. Figure 10 As shown, the ultrasound image analysis method 1000 of this embodiment includes the following steps:

[0234] Step S1010: Obtain an ultrasound image by performing an ultrasound scan on the breast region of the subject.

[0235] Step S1020: Detect lesions in the ultrasound image;

[0236] Step S1030: Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types;

[0237] Step S1040: Determine the BI-RADS grade of the lesion based on the pre-trained BI-RADS grading model;

[0238] Step S1050: Generate a radar chart using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one BI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing a score.

[0239] In step S1060, a feature graphic is generated on the radar chart based on the BI-RADS score corresponding to the determined BI-RADS feature type;

[0240] Step S1070: Display the radar chart, the feature graph, and the BI-RADS classification.

[0241] The ultrasound image analysis method 1000 of this invention can be implemented by the ultrasound imaging system described above. The relevant descriptions of each step can be referred to the relevant descriptions above, and will not be repeated here.

[0242] According to the ultrasound image analysis method 900 and ultrasound imaging system of the present invention, the BI-RADS score under each BI-RADS feature type of breast lesion is quantified, and the quantified BI-RADS is presented intuitively through radar charts and feature graphs. Simultaneously, the BI-RADS grading of the lesion is compared with the radar... Figure 1 The same display is helpful in guiding and optimizing the analysis of lesions in ultrasound images.

[0243] This application also provides an ultrasound imaging system. (Continuing to refer to...) Figure 1 The ultrasound imaging system includes an ultrasound probe 110, a transmitting circuit 112, a receiving circuit 114, a processor 116, and a display 118. The relevant descriptions of each component can be found in the description of the ultrasound imaging system 100 above. The following only describes the main functions of the ultrasound imaging system and omits the details already described above.

[0244] Specifically, the transmitting circuit 112 is used to excite the ultrasound probe 110 to emit ultrasound waves toward the breast region of the subject; the receiving circuit 114 is used to control the ultrasound probe 110 to receive the ultrasound echo returned from the breast region to obtain an ultrasound echo signal; the processor 116 is used to: acquire an ultrasound image obtained by performing an ultrasound scan on the breast region of the subject; detect lesions in the ultrasound image; determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; generate a radar chart using the at least seven BI-RADS feature types as classification axes of the radar chart; wherein the classification axis divides the radar chart into multiple partitions, each classification axis or each partition is used to represent a BI-RADS feature type, and at least one classification axis or at least one partition has a scale unit for representing the score; generate a feature graph on the radar chart based on the BI-RADS score corresponding to the determined BI-RADS feature type; and the display 118 is used to display the radar chart and the feature graph.

[0245] The ultrasound imaging system in this embodiment is largely similar to the ultrasound imaging system described above; some identical details are omitted below. In this embodiment, the BI-RADS score for each BI-RADS feature type of the lesion is quantified, and a feature graph is plotted on a radar chart based on the BI-RADS score of the lesion under each BI-RADS feature type. The specific forms of the radar chart and feature graph can be found above. Unlike the ultrasound imaging system described above, this embodiment does not limit the determination of the BI-RADS grade of the lesion, nor does it limit the correlation between the BI-RADS grade and the radar chart. Figure 1 Similarly, it is only necessary to present the BI-RADS score of the lesion under each BI-RADS feature type through a radar chart.

[0246] In one embodiment, the BI-RADS score for each BI-RADS feature type of a lesion is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features under each BI-RADS feature type. In another embodiment, the BI-RADS score for each BI-RADS feature type of a lesion is the contribution of the BI-RADS lesion features under each BI-RADS feature type to the BI-RADS classification.

[0247] The following is a reference to the appendix. Figure 11 A method for analyzing ultrasound images according to another embodiment of this application is described. Figure 11 A schematic flowchart of an ultrasound image analysis method 1100 according to another embodiment of this application is shown. Figure 11 As shown, the ultrasound image analysis method 1100 includes the following steps:

[0248] In step S1110, an ultrasound image is obtained by performing an ultrasound scan on the breast region of the subject.

[0249] In step S1120, lesions in the ultrasound image are detected;

[0250] In step S1130, the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types is determined;

[0251] In step S1140, a radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein, the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score;

[0252] In step S1150, a feature graphic is generated on the radar chart based on the BI-RADS score corresponding to the determined BI-RADS feature type;

[0253] In step S1160, the radar image and the feature pattern are displayed.

[0254] According to the ultrasound image analysis method 1100 and ultrasound imaging system of this application, the BI-RADS score of each BI-RADS feature type of breast lesion is quantified, and the quantified BI-RADS is presented intuitively through radar chart and feature graph, which is helpful to guide and optimize the analysis of lesions in ultrasound images.

[0255] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

[0256] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0257] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0258] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0259] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0260] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or elements of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature serving the same, equivalent, or similar purpose.

[0261] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0262] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing some or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0263] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. An ultrasound imaging system, characterized in that, The ultrasound imaging system includes: Ultrasonic probe; A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the thyroid region of the subject being tested; A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the thyroid region to obtain ultrasound echo signals; Processor, used for: Acquire ultrasound images by scanning the thyroid region of the subject. Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; Determine the TI-RADS score corresponding to the TI-RADS lesion characteristics, and determine the TI-RADS grade of the lesion based on the sum of the TI-RADS scores; A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the TI-RADS score corresponding to the determined TI-RADS lesion features, a feature graphic is generated on the radar image. The color of the feature graphic is determined according to the TI-RADS grade of the lesion or according to the sum of the TI-RADS scores. Different colors of the feature graphic represent different TI-RADS grade results. Furthermore, the parts of the feature graphic corresponding to different TI-RADS scores are displayed as different shades of the color. A display for showing the radar chart, the feature graph, and the TI-RADS classification.

2. The ultrasound imaging system according to claim 1, characterized in that, The TI-RADS feature types include shape type, composition type, echo type, calcification type, and edge type.

3. The ultrasound imaging system according to claim 1, characterized in that, The radar chart is a circular radar chart, and each partition is used to represent a TI-RADS feature type. The area of ​​the feature graphic within each partition represents the TI-RADS score of the TI-RADS feature type corresponding to the partition. Each partition is identified with the TI-RADS feature type corresponding to the partition.

4. The ultrasound imaging system according to claim 1, characterized in that, The radar chart is a polygon radar chart, and each classification axis is used to represent a TI-RADS feature type. The feature graph is a graph formed by connecting the coordinate points of the TI-RADS score representing the TI-RADS feature type corresponding to the classification axis on each classification axis. Each classification axis is marked with the TI-RADS feature type corresponding to the classification axis.

5. The ultrasound imaging system according to claim 4, characterized in that, The origin of the coordinate system on the classification axis, representing a TI-RADS score of 0, is located away from the center of the radar chart.

6. The ultrasound imaging system according to claim 1, characterized in that, At least one of the partitions has grid lines that divide the partition into at least two sub-intervals.

7. The ultrasound imaging system according to claim 1, characterized in that, The first classification axis in the classification axis includes a maximum scale unit, a minimum scale unit, and a preset scale unit, wherein the preset scale unit is used to represent the maximum TI-RADS score that can be obtained for each TI-RADS lesion feature under the first TI-RADS feature type corresponding to the first classification axis; the processor is further configured to: When the preset scale unit is smaller than the maximum scale unit but larger than the minimum scale unit, the display is controlled to differentiate the portion between the maximum scale unit and the preset scale unit in the first classification axis from the portion between the minimum scale unit and the preset scale unit.

8. The ultrasound imaging system according to claim 1, characterized in that, The processor is also configured to: determine the color of the radar map based on the TI-RADS grading of the lesion or based on the sum of the TI-RADS scores.

9. The ultrasound imaging system according to claim 1, characterized in that, The radar chart also includes a base map, and the classification axis extends from the central region of the base map to the edge of the base map, dividing the base map into multiple partitions.

10. The ultrasound imaging system according to claim 9, characterized in that, The base map is a circle or a polygon, and the number of sides of the polygon is equal to the number of TI-RADS feature types.

11. The ultrasound imaging system according to claim 9, characterized in that, The processor is also used for: Calculate the ratio of the area of ​​the feature graphic to the area of ​​the base image, and control the display to show the ratio.

12. The ultrasound imaging system according to claim 11, characterized in that, The ratio is displayed around the radar image, or the ratio is displayed around the lesion in the ultrasound image.

13. The ultrasound imaging system according to claim 1, characterized in that, The processor is further configured to: control the display to simultaneously display the radar chart or the feature graphic with at least one of the following: The ultrasound image, thyroid position diagram, region of interest in the ultrasound image, boundary of lesion detected in the ultrasound image, and personal information of the subject being tested.

14. The ultrasound imaging system according to claim 13, characterized in that, The processor is also used for: When it is determined that the user selects a position in the radar chart or feature graph corresponding to each TI-RADS feature type, the display is controlled to show the effect corresponding to the TI-RADS feature type in the ultrasound image.

15. The ultrasound imaging system according to claim 14, characterized in that, When it is determined that the user selects a position in the radar chart or feature graph corresponding to each TI-RADS feature type, the control of the display to show an effect in the ultrasound image that corresponds to the TI-RADS feature type includes at least one of the following: When it is determined that the user selects a position in the radar image or the feature graphic that corresponds to the shape type and / or edge type, the display is controlled to highlight the boundary of the lesion area in the ultrasound image; When it is determined that the user selects a position in the radar map or the feature graph that corresponds to the echo type and / or component type, the display is controlled to highlight the lesion area in the ultrasound image; When it is determined that the user selects a location in the radar image or feature graph that corresponds to the type of calcification, the display is controlled to highlight the calcification area extracted from the ultrasound image.

16. The ultrasound imaging system according to claim 1, characterized in that, The processor is used to identify the TI-RADS lesion features under each TI-RADS feature type in at least one of the following ways: The processor is used to predict TI-RADS lesion features under multiple TI-RADS feature types based on a multi-task neural network model. The processor is used to predict TI-RADS lesion features under a single TI-RADS feature type based on a single-task neural network model. The processor is used to extract image features for a single TI-RADS feature type using a feature extraction algorithm, and to classify the extracted image features to obtain TI-RADS lesion features under the TI-RADS feature type.

17. The ultrasound imaging system according to claim 1, characterized in that, The processor is used to determine the TI-RADS grade of the lesion based on the TI-RADS score, including: The processor is used to sum the TI-RADS scores to obtain a total TI-RADS score; and to determine the TI-RADS grade of the lesion based on the correspondence between the total TI-RADS score and the TI-RADS grade.

18. An ultrasound imaging system, characterized in that, The ultrasound imaging system includes: Ultrasonic probe; A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the thyroid region of the subject being tested; A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the thyroid region to obtain ultrasound echo signals; Processor, used for: Acquire ultrasound images by scanning the thyroid region of the subject. Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; Determine the TI-RADS score corresponding to the TI-RADS lesion features; A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the TI-RADS scores corresponding to the determined TI-RADS lesion features, a feature graphic is generated on the radar image. The color of the feature graphic is determined according to the sum of the TI-RADS scores. Different colors of the feature graphic represent different TI-RADS grading results. Furthermore, the portions of the feature graphic corresponding to different TI-RADS scores are displayed as different shades of the color. A display for showing the radar image and the feature graph.

19. The ultrasound imaging system according to claim 18, characterized in that, The processor is also used for: Control the display to show the sum of all the TI-RADS scores.

20. The ultrasound imaging system according to claim 18, characterized in that, The radar chart also includes a base map, and the classification axis extends from the central region of the base map to the edge of the base map to divide the base map into multiple partitions. The processor is further configured to: Calculate the ratio of the area of ​​the feature graphic to the area of ​​the base image, and control the display to show the ratio.

21. A method for analyzing ultrasound images, characterized in that, The method includes: Acquire ultrasound images by scanning the thyroid region of the subject. Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; Determine the TI-RADS score corresponding to the TI-RADS lesion characteristics, and determine the TI-RADS grade of the lesion based on the sum of the TI-RADS scores; A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the TI-RADS score corresponding to the determined TI-RADS lesion features, a feature graphic is generated on the radar image. The color of the feature graphic is determined according to the TI-RADS grade of the lesion or according to the sum of the TI-RADS scores. Furthermore, the portions of the feature graphic corresponding to different TI-RADS scores are displayed as different shades of the color. The radar chart, the feature graph, and the TI-RADS classification are displayed.

22. An ultrasound image analysis method, characterized in that, The method includes: Acquire ultrasound images by scanning the thyroid region of the subject. Detect lesions in the ultrasound images and identify the TI-RADS lesion features corresponding to the lesions under at least five TI-RADS feature types; Determine the TI-RADS score corresponding to the TI-RADS lesion features; A radar chart is generated using the at least five TI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the TI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the TI-RADS scores corresponding to the determined TI-RADS lesion features, a feature graphic is generated on the radar image. The color of the feature graphic is determined according to the sum of the TI-RADS scores. Different colors of the feature graphic represent different TI-RADS grading results. Furthermore, the portions of the feature graphic corresponding to different TI-RADS scores are displayed as different shades of the color. The radar image and the feature graph are displayed.

23. An ultrasound imaging system, characterized in that, The ultrasound imaging system includes: Ultrasonic probe; A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the breast region of the subject being tested; A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the breast region to obtain ultrasound echo signals; Processor, used for: Acquire ultrasound images by scanning the breast region of the subject. Detecting lesions in the ultrasound images; Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; The BI-RADS classification of the lesion was determined based on a pre-trained BI-RADS classification model; A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the BI-RADS score corresponding to the determined BI-RADS feature type, a feature graphic is generated on the radar chart. The color of the feature graphic is determined according to the BI-RADS grade of the lesion or according to the sum of the BI-RADS scores. Different colors of the feature graphic represent different BI-RADS grade results. Furthermore, the parts of the feature graphic corresponding to different BI-RADS scores are displayed as different shades of the color. A display for showing the radar chart, the feature graph, and the BI-RADS classification.

24. The ultrasound imaging system according to claim 23, characterized in that, The BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types, or... The BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types is the contribution of the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types to the BI-RADS classification.

25. The ultrasound imaging system according to claim 24, characterized in that, When the BI-RADS score represents the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types, the processor determines the BI-RADS score corresponding to each BI-RADS feature type among the at least seven BI-RADS feature types in at least one of the following ways: The processor is used to input the ultrasound images into a pre-trained deep learning model for each BI-RADS feature type, and the probability prediction branch of the deep learning model predicts the probability under each BI-RADS feature type, and the feature prediction branch of the deep learning model predicts the BI-RADS lesion features under each BI-RADS feature type. The processor is used to extract image features of the ultrasound image for each BI-RADS feature type based on a feature extraction algorithm, and based on the image features, to predict the probability under each BI-RADS feature type using a pre-trained probability prediction machine learning model, and to predict the BI-RADS lesion features corresponding to each BI-RADS feature type using a pre-trained feature prediction machine learning model. The processor is used to extract image features of the ultrasound image using a deep learning model and a feature extraction algorithm, respectively, and to fuse the image features extracted by the deep learning model and the feature extraction algorithm to obtain fused image features corresponding to each BI-RADS feature type, and to predict the probability based on the fused image features.

26. The ultrasound imaging system according to claim 24, characterized in that, When the BI-RADS score is the contribution of the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types to the BI-RADS classification, the processor determines the BI-RADS score corresponding to each of the at least seven BI-RADS feature types of the lesion, including: The processor is used to extract image features of the ultrasound image under each BI-RADS feature type, and integrate the image features under each BI-RADS feature type into a high-dimensional feature; Predict the BI-RADS classification based on the high-dimensional features; Based on the image features under each BI-RADS feature type, a machine learning model is used to evaluate the contribution of BI-RADS lesion features under each BI-RADS feature type to the BI-RADS classification.

27. The ultrasound imaging system according to claim 23, characterized in that, The BI-RADS feature types include orientation type, shape type, edge type, internal echo type, posterior echo type, calcification type, and blood flow type.

28. The ultrasound imaging system according to claim 23, characterized in that, The radar chart is a circular radar chart, and each partition is used to represent a BI-RADS feature type. The area of ​​the feature graphic within each partition represents the BI-RADS score of the BI-RADS feature type corresponding to the partition. Each partition is identified with the BI-RADS feature type corresponding to the partition.

29. The ultrasound imaging system according to claim 23, characterized in that, The radar chart is a polygon radar chart, and each classification axis is used to represent a BI-RADS feature type. The feature graph is a graph formed by connecting the coordinate points of the BI-RADS score representing the BI-RADS feature type corresponding to each classification axis. Each classification axis is marked with the BI-RADS feature type corresponding to the classification axis.

30. The ultrasound imaging system according to claim 23, characterized in that, At least one of the partitions has grid lines that divide the partition into at least two sub-intervals.

31. The ultrasound imaging system according to claim 23, characterized in that, The processor is also configured to: determine the color of the radar image based on the BI-RADS classification of the lesion.

32. The ultrasound imaging system according to claim 23, characterized in that, The radar chart also includes a base map, and the classification axis extends from the central region of the base map to the edge of the base map to divide the base map into multiple partitions.

33. The ultrasound imaging system according to claim 32, characterized in that, The base map is a circle or a polygon, and the number of sides of the polygon is equal to the number of BI-RADS feature types.

34. The ultrasound imaging system according to claim 32, characterized in that, The BI-RADS score corresponding to at least seven BI-RADS feature types of the lesion is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types, and the processor is further configured to: Calculate the ratio of the area of ​​the feature graphic to the area of ​​the base image, and control the display to show the ratio.

35. The ultrasound imaging system according to claim 23, characterized in that, The processor is further configured to: control the display to simultaneously display the radar chart or the feature graphic with at least one of the following: The ultrasound image, breast position map, region of interest in the ultrasound image, boundary of lesion detected in the ultrasound image, and personal information of the subject being tested.

36. The ultrasound imaging system according to claim 35, characterized in that, The processor is also used for: When it is determined that the user selects a position in the radar chart or feature graph corresponding to each BI-RADS feature type, the display is controlled to show the effect corresponding to the BI-RADS feature type in the ultrasound image.

37. The ultrasound imaging system according to claim 36, characterized in that, When it is determined that the user has selected a position in the radar chart or feature graph corresponding to each BI-RADS feature type, the control of the display to show an effect in the ultrasound image that corresponds to the BI-RADS feature type includes at least one of the following: When it is determined that the user selects a position in the radar image or the feature graphic that corresponds to the shape type and / or edge type, the display is controlled to highlight the boundary of the lesion area in the ultrasound image; When it is determined that the user selects a position in the radar map or the feature graphic that corresponds to the edge type, internal echo type and / or posterior echo type, the display is controlled to highlight the lesion area in the ultrasound image; When it is determined that the user selects a position in the radar map or the feature graph that corresponds to the type of calcification, the display is controlled to highlight the calcification area extracted from the ultrasound image. When it is determined that the user selects a position in the radar image or the feature graphic that corresponds to the blood flow type, a color blood flow image corresponding to the ultrasound image is extracted, and the display is controlled to overlay the color blood flow image onto the ultrasound image.

38. An ultrasound image analysis method, characterized in that, The method includes: Acquire ultrasound images by scanning the breast region of the subject. Detecting lesions in the ultrasound images; Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; The BI-RADS classification of the lesion was determined based on a pre-trained BI-RADS classification model; A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the BI-RADS score corresponding to the determined BI-RADS feature type, a feature graphic is generated on the radar chart. The color of the feature graphic is determined according to the BI-RADS grade of the lesion or according to the sum of the BI-RADS scores. Different colors of the feature graphic represent different BI-RADS grade results. Furthermore, the parts of the feature graphic corresponding to different BI-RADS scores are displayed as different shades of the color. The radar chart, the feature graph, and the BI-RADS classification are displayed.

39. An ultrasound imaging system, characterized in that, The ultrasound imaging system includes: Ultrasonic probe; A transmitting circuit is used to excite the ultrasound probe to emit ultrasound waves toward the breast region of the subject being tested; A receiving circuit is used to control the ultrasound probe to receive ultrasound echoes returned from the breast region to obtain ultrasound echo signals; Processor, used for: Acquire ultrasound images by scanning the breast region of the subject. Detecting lesions in the ultrasound images; Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graphic is generated on the radar chart. The color of the feature graphic is determined according to the sum of the BI-RADS scores. Different colors of the feature graphic represent different BI-RADS rating results. Furthermore, the parts of the feature graphic corresponding to different BI-RADS scores are displayed as different shades of the color. A display for showing the radar image and the feature graph.

40. The ultrasound imaging system according to claim 39, characterized in that, The BI-RADS score corresponding to at least seven BI-RADS feature types of the lesion is the probability that the lesion is indicated as a preset lesion state by the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types, or, The BI-RADS score corresponding to at least seven BI-RADS feature types of the lesion is the contribution of the BI-RADS lesion features corresponding to the at least seven BI-RADS feature types to the BI-RADS classification of the lesion.

41. A method for analyzing ultrasound images, characterized in that, The method includes: Acquire ultrasound images by scanning the breast region of the subject. Detecting lesions in the ultrasound images; Determine the BI-RADS score corresponding to the lesion under at least seven BI-RADS feature types; A radar chart is generated using the at least seven BI-RADS feature types as classification axes; wherein the classification axes divide the radar chart into multiple partitions, each classification axis or each partition is used to represent one of the BI-RADS feature types, and at least one classification axis or at least one partition has a scale unit for representing a score; Based on the determined BI-RADS feature type and the corresponding BI-RADS score, a feature graphic is generated on the radar chart. The color of the feature graphic is determined according to the sum of the BI-RADS scores. Different colors of the feature graphic represent different BI-RADS rating results. Furthermore, the parts of the feature graphic corresponding to different BI-RADS scores are displayed as different shades of the color. The radar image and the feature graph are displayed.

Images