Medical image tumor detection method and device, computer equipment and storage medium

Abstract

The invention relates to a medical image tumor detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring an ultrasonic video image and a contrast image of a detection part; the ultrasonic video image is input into a Mem-SAM2 model based on a double-layer memory library mechanism, the Mem-SAM2 model based on the double-layer memory library mechanism adopts a short-term memory library and a long-term memory library to segment the ultrasonic video image, and a segmentation label of each ultrasonic video frame is generated; superposing the segmentation label of each ultrasonic video frame onto a corresponding angiography image frame, and segmenting each angiography image frame into a tumor area and a background area; performing time sequence intensity curve extraction on the tumor area and the background area to obtain time sequence characteristics of the ultrasonic video image; and performing tumor benign and malignant detection according to the time sequence characteristics. According to the method, the existing Mem-SAM2 model is optimized, so that the reliability and consistency of image segmentation are improved, and the segmentation accuracy and robustness of the ultrasonic video image are improved.

Description

Technical Field

[0001] This application belongs to the technical field of biomedical image processing, and particularly relates to a medical image tumor detection method, device, computer device, and storage medium. Background Art

[0002] Breast cancer is one of the most common malignant tumors among women globally. Early diagnosis is crucial for improving the treatment effect and survival rate. Imaging screening of breast tumors plays a key role in the early detection, diagnosis, and treatment of breast cancer. As a non-invasive, non-radiative, and low-cost examination method, ultrasound (US) plays an important role in the preliminary screening of breast tumors. However, US has limitations in tumor invasiveness and benign / malignant differentiation, mainly including: (1) It only provides anatomical structure information and is difficult to reflect the blood supply characteristics of lesions; (2) It has limited diagnostic ability for lesions lacking typical imaging features and is easily affected by the operator's experience; (3) It is easily interfered by artifacts in some lesions (such as highly calcified lesions), affecting the judgment of the boundary and internal echo. Contrast Enhanced Ultrasound (CEUS) overcomes the deficiencies of US in tumor benign / malignant judgment by observing the microvascular perfusion pattern in real time and dynamically.

[0003] Medical image automatic segmentation technology plays a crucial role in clinical diagnosis, especially in the analysis of ultrasound images and other dynamic image data. The introduction of automation technology has greatly improved the diagnostic efficiency and accuracy. However, the ultrasound image data volume of breast tumors is huge, especially the number of frames in ultrasound videos often reaches several thousand or even tens of thousands. Traditional manual segmentation methods require doctors to accurately label the tumor area frame by frame, which is extremely laborious and easily affected by the operator's subjective factors. Especially in high-frame-rate ultrasound videos, with a large number of frames and fast-changing images, frame-by-frame segmentation is not only inefficient but also difficult to ensure the accuracy of segmentation.

[0004] Currently, Segment Anything Model 2 (SAM2) has been widely used in medical video automatic segmentation tasks. Different from the traditional SAM model, SAM2 is optimized for video data and can process dynamic images through video object tracking. By this method, SAM2 can perform efficient automatic segmentation by utilizing the relationship between consecutive frames. However, SAM2 has certain limitations when directly applied to ultrasound video segmentation. Especially in ultrasound videos, due to its strong noise interference and the complexity of temporal information, it is often difficult to obtain high-quality segmentation results solely relying on the method of SAM2.

[0005] To address these deficiencies, researchers have proposed the MedSAM-2 framework based on SAM2. MedSAM-2 introduces a self-sorting memory bank mechanism that can dynamically select more informative embedding features based on confidence and similarity, regardless of the order of the time series. This not only significantly improves performance in 3D medical image segmentation but also provides effective support for single-shot prompt segmentation of 2D images. Even without explicitly considering the temporal relationship between frames, efficient segmentation can still be achieved. However, although MedSAM-2 has achieved certain success in medical image segmentation, its application scope is relatively wide, and it has not deeply solved the problem of accurate segmentation in specific fields such as breast ultrasound images. When faced with specific medical images such as breast ultrasound, it may be difficult to effectively capture the detailed features of the image and the deep association with pathological information. In addition, since MedSAM-2 relies on a single memory bank to store all feature representations, this may lead to redundant information storage and difficulty in accurately capturing the details of key frames in specific tasks. On the other hand, in existing tumor benign and malignant classification tasks, traditional methods often rely on manual selection of key frames, which results in incomplete and inaccurate extraction of specific features by the model and affects the final classification effect. Summary of the Invention

[0006] The present application provides a method, device, computer device, and storage medium for tumor detection in medical images, aiming to at least partly solve one of the above technical problems in the prior art.

[0007] To solve the above problems, the present application provides the following technical solutions:

[0008] A method for tumor detection in medical images, comprising:

[0009] Obtaining ultrasound video images and contrast images of the detection site;

[0010] Inputting the ultrasound video images into the Mem-SAM2 model based on a double-layer memory bank mechanism, and the Mem-SAM2 model based on a double-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasound video images, generating segmentation labels for each ultrasound video frame;

[0011] Overlaying the segmentation labels of each ultrasound video frame onto the corresponding contrast image frame respectively, and segmenting each contrast image frame into a tumor region and a background region;

[0012] Performing temporal intensity curve extraction on the tumor region and the background region to obtain the temporal features of the ultrasound video images;

[0013] Performing tumor benign and malignant detection based on the temporal features.

[0014] The technical solutions adopted in the embodiments of this application further include: The obtaining of the ultrasonic video image and the contrast image of the detection site further includes:

[0015] Performing radiomics feature extraction on the ultrasonic video image; the radiomics features include organ size, shape, structure, metabolism, blood flow, or / and intensity features.

[0016] The technical solutions adopted in the embodiments of this application further include: Before inputting the ultrasonic video image into the Mem-SAM2 model based on the double-layer memory bank mechanism and using the short-term memory bank and the long-term memory bank of the Mem-SAM2 model based on the double-layer memory bank mechanism to segment the ultrasonic video image, it further includes:

[0017] Performing preprocessing on the ultrasonic video image, and the preprocessing includes normalization and standardization processing.

[0018] The technical solutions adopted in the embodiments of this application further include: The Mem-SAM2 model based on the double-layer memory bank mechanism includes a short-term memory bank and a long-term memory bank. The short-term memory bank is used to store the key information at the current time point, and the key information includes the current video frame and the video frames similar to the current video frame; when the short-term memory bank is full, transfer the video frame with the highest quality and the highest similarity to the current video frame to the long-term memory bank for storage, discard the video frame with the lowest similarity to the current video frame, and then store a new video frame.

[0019] The technical solutions adopted in the embodiments of this application further include: The extracting of the time-intensity curve for the tumor region and the background region to obtain the time series features of the ultrasonic video image is specifically:

[0020] The time-intensity curve is used to reflect the brightness changes of the tumor region and the background region on the time axis, analyze the wash-in and wash-out characteristics reflected by the time-intensity curve and the brightness difference between the tumor region and the background region during the contrast period, and extract the time series features that can reflect the benign and malignant nature of the tumor from the ultrasonic video image according to the analysis result of the brightness difference.

[0021] The technical solutions adopted in the embodiments of this application further include: The performing of the benign and malignant detection of the tumor according to the time series features is specifically:

[0022] Inputting the radiomics features and the time series features into a classification model together, and outputting the benign and malignant classification result of the tumor through the classification model.

[0023] The technical solutions adopted in the embodiments of this application further include: The classification model adopts a support vector machine or a random forest classifier.

[0024] Another technical solution adopted in the embodiments of the present application is: A medical imaging tumor detection device, comprising:

[0025] An image acquisition module: configured to acquire ultrasonic video images and contrast images of the detection site;

[0026] A first image segmentation module: configured to input the ultrasonic video images into the Mem-SAM2 model based on a double-layer memory bank mechanism, and the Mem-SAM2 model based on the double-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasonic video images to generate segmentation labels for each ultrasonic video frame;

[0027] A second image segmentation module: configured to respectively superimpose the segmentation labels of each ultrasonic video frame onto the corresponding contrast image frame, and segment each contrast image frame into a tumor region and a background region;

[0028] A feature extraction module: configured to perform temporal intensity curve extraction on the tumor region and the background region to obtain the temporal features of the ultrasonic video images;

[0029] A tumor detection module: configured to perform benign and malignant tumor detection based on the temporal features.

[0030] Another technical solution adopted in the embodiments of the present application is: A computer device, the computer device includes a processor and a memory coupled to the processor, wherein,

[0031] The memory stores program instructions for implementing the medical imaging tumor detection method;

[0032] The processor is configured to execute the program instructions stored in the memory to control the medical imaging tumor detection method.

[0033] Another technical solution adopted in the embodiments of the present application is: A storage medium stores program instructions executable by a processor, and the program instructions are used to execute the medical imaging tumor detection method.

[0034] Compared with the prior art, the beneficial effects produced by the embodiments of the present application are as follows: The medical image tumor detection method, device, computer device, and storage medium of the embodiments of the present application optimize the existing Mem-SAM2 model by introducing a short-term memory bank and a long-term memory bank. The optimized Mem-SAM2 model is used to automatically segment ultrasound video images to generate a segmentation mask for each ultrasound video frame, and then the segmentation mask is superimposed on the corresponding contrast image frame to segment each contrast image frame into a tumor region and a background region, making the segmentation process more standardized and automated, avoiding the interference of human factors on the segmentation results, thereby improving the reliability and consistency of image segmentation, and improving the segmentation accuracy and robustness of ultrasound video images. And based on the segmentation results, the time-intensity curves of the tumor region and the background region are further analyzed to extract temporal features that can reflect the benign and malignant nature of the tumor. The temporal features are combined with radiomics features for tumor benign and malignant classification, enhancing the accuracy and interpretability of the tumor classification results, and enhancing the stability and adaptability of the classification model, so as to provide more efficient and reliable diagnostic support in clinical practice and have broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flowchart of the medical image tumor detection method according to the embodiment of the present application;

[0036] Figure 2 is a schematic diagram of the framework of the medical image tumor detection system according to the embodiment of the present application;

[0037] Figure 3 is a schematic structural diagram of the medical image tumor detection device according to the embodiment of the present application;

[0038] Figure 4 is a schematic structural diagram of the computer device according to the embodiment of the present application;

[0039] Figure 5 is a schematic structural diagram of the storage medium according to the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0041] The terms "first", "second", and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", and "third" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined. In the embodiments of this application, all directional indications (such as up, down, left, right, front, back...) are only used to explain the relative positional relationship, movement conditions, etc. between components in a specific posture (as shown in the drawings). If the specific posture changes, then the directional indications will also change accordingly. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or computer device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or computer devices.

[0042] Reference to "embodiment" in this context means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0043] Specifically, please refer to Figure 1 and Figure 2 , Figure 1 is a flowchart of the medical image tumor detection method according to the embodiment of this application, Figure 2 is a schematic diagram of the framework of the medical image tumor detection system according to the embodiment of this application. The medical image tumor detection method according to the embodiment of this application includes the following steps:

[0044] S100: Obtain the ultrasonic video image and the contrast image of the detection site, and extract the radiomics features from the ultrasonic video image;

[0045] In this step, the ultrasonic video image includes, but is not limited to, various types of medical images such as B-ultrasound, CT / MRI, and colonoscopy videos of various lesion sites such as the breast, thyroid, lung, or liver. The radiomics features can capture the minute changes in the ultrasonic video image and are obtained by extracting the high-dimensional data in the ultrasonic video image, including morphological features such as organ size, shape, or / and structure, and functional features such as metabolism, blood flow, or / and intensity. Extracting the radiomics features can help clinicians understand the physiological and pathological mechanisms of diseases and provide strong support for disease assessment and treatment, etc.

[0046] S110: Input the ultrasonic video image into the Mem-SAM2 model based on the dual-layer memory bank mechanism. The Mem-SAM2 model based on the dual-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasonic video image, generating a segmentation mask (label) for each ultrasonic video frame.

[0047] In this step, first, preprocess the ultrasonic video image, such as standardization and normalization, and then input it into the Mem-SAM2 model based on the dual-layer memory bank mechanism. The Mem-SAM2 model based on the dual-layer memory bank mechanism uses an automatic segmentation algorithm to automatically segment the ultrasonic video image, generating a segmentation mask for each frame of the ultrasonic image. Specifically, the Mem-SAM2 model with the dual-layer memory bank mechanism has been innovatively optimized on the basis of the existing Mem-SAM2 network. By introducing a short-term memory bank and a long-term memory bank, the short-term memory bank is used to store the key information at the current time point. The key information includes the current video frame and the video frames similar to the current video frame. When the memory bank is full, transfer the video frame with the highest quality and the highest similarity to the current video frame to the long-term memory bank. After discarding the frame with the lowest similarity to the current video frame, store a new video frame, so as to continuously create new storage space for the short-term memory bank to ensure that the model can retain important historical information. The long-term memory bank is used to store the high-quality video frames screened by the short-term memory bank, thereby improving the segmentation accuracy and robustness of the ultrasonic video image.

[0048] It can be understood that through the collaborative effect of the two memory banks in the embodiment of the present application, when processing the ultrasonic video image, the memory level is dynamically adjusted according to the actual contribution of the video frame. It can effectively capture and utilize the detailed features of the ultrasonic video image and perform automatic segmentation on the premise of ensuring non-redundant information storage, making the segmentation process more standardized and automated, avoiding the interference of human factors on the segmentation result, thereby improving the reliability and consistency of image segmentation, and can effectively cope with the diversity of different ultrasonic devices, operators, and imaging qualities, improving the universality and stability of the model. By automatically segmenting the ultrasonic video image through the automatic segmentation algorithm, the model can accurately process the details in the ultrasonic video image, especially when dealing with the image changes in the time series, it can effectively avoid information loss or redundancy.

[0049] Furthermore, in order to further optimize the segmentation accuracy and the ability to process long time-series data, the present application can also use the Mamba architecture based on the State Space Model (abbreviated as SSM) as an alternative to the Mem-SAM2 model based on the double-layer memory bank mechanism. The Mamba architecture adopts an improved Temporal Mamba Block, which can effectively compress long spatio-temporal representations, expand the receptive field, improve the application effect in medical videos, efficiently model long sequence dependencies, show higher computational efficiency when processing long time-series, further improve the efficiency and robustness of image segmentation, achieve a more efficient automatic segmentation process, and ensure higher segmentation accuracy at the same time.

[0050] S120: Respectively superimpose the segmentation masks of each ultrasound video frame onto the corresponding contrast-enhanced ultrasound (CEUS) image frame, and segment each CEUS image frame into a tumor region and a background region;

[0051] In this step, by superimposing the segmentation mask of the ultrasound video frame onto the corresponding CEUS image frame, the CEUS image frame is automatically segmented into a tumor region and a background region without relying on human factors, eliminating the subjective errors and biases brought by manual segmentation, ensuring the consistency and accuracy of the segmentation results. Especially when dealing with large-scale clinical data, it can significantly improve work efficiency.

[0052] S130: Extract the time-intensity curve (abbreviated as TIC) for the segmented tumor region and background region to obtain temporal features that can reflect the benign or malignant nature of the tumor;

[0053] In this step, combined with clinical analysis, it can be known that when the brightness of the tumor region in CEUS is higher than that of the background region, it usually indicates rich blood flow and may be a malignant tumor; while when the brightness of the tumor region in CEUS is lower than the background, it may be calcification or other breast diseases; if the brightness difference between the tumor region and the background region in CEUS is not significant, it usually indicates that it may be a benign tumor. Therefore, based on the segmented tumor region and background region, the present application further extracts the TIC curve for the tumor region and background region. The TIC curve reflects the brightness changes of the tumor region and background region on the time axis. By analyzing the wash-in and wash-out characteristics reflected by the TIC curve and the brightness difference between the tumor region and the background region during CEUS, temporal features that can reflect the benign or malignant nature of the tumor can be effectively extracted from the ultrasound video images, improving the extraction depth and accuracy of tumor features, ensuring that the classification model can make more accurate judgments from comprehensive and accurate data, avoiding the simple processing or empirical judgment of ultrasound video images by traditional methods, and providing a more objective and accurate basis for the clinical judgment of the benign or malignant nature of the tumor.

[0054] In addition, by comparing the temporal features of the foreground and background parts of a single CEUS image, the present application can significantly reduce the imaging differences caused by factors such as different ultrasonic equipment models and differences in the habits of operating technicians, thereby making the model more universal and robust.

[0055] S140: Input the extracted radiomics features and temporal features into a classification model, and output the classification result of tumor benignity and malignancy through the classification model;

[0056] In this step, the radiomics features are obtained by extracting high-dimensional data from ultrasonic video images and can capture the subtle changes in ultrasonic video images. Therefore, on the basis of temporal features, the present application further combines radiomics features for tumor benignity and malignancy classification to enhance the accuracy and interpretability of tumor classification results, and enhance the stability and adaptability of the classification model, so as to provide more efficient and reliable diagnostic support in clinical practice and has broad application prospects.

[0057] Furthermore, in order to complete the tumor benignity and malignancy classification task, the present application uses machine learning classifiers such as support vector machines or random forests to train and predict classification using the extracted radiomics features and temporal features. Through the combination and analysis of multi-dimensional features, it can be applied to a wider range of medical image analysis scenarios and can show good application potential in complex clinical diagnosis environments.

[0058] Based on the above, the medical image tumor detection method of the embodiment of the present application optimizes the existing Mem-SAM2 model by introducing a short-term memory bank and a long-term memory bank, uses the optimized Mem-SAM2 model to automatically segment ultrasonic video images, generates a segmentation mask for each ultrasonic video frame, and then superimposes the segmentation mask on the corresponding contrast image frame to segment each contrast image frame into a tumor region and a background region, making the segmentation process more standardized and automated, avoiding the interference of human factors on the segmentation result, thereby improving the reliability and consistency of image segmentation, and improving the segmentation accuracy and robustness of ultrasonic video images. And based on the segmentation result, further analyze the time-intensity curves of the tumor region and the background region, extract temporal features that can reflect tumor benignity and malignancy, combine the temporal features with radiomics features for tumor benignity and malignancy classification, enhance the accuracy and interpretability of tumor classification results, and enhance the stability and adaptability of the classification model, so as to provide more efficient and reliable diagnostic support in clinical practice and has broad application prospects.

[0059] Please refer to Figure 3 , which is a schematic structural diagram of the medical image tumor detection device of the embodiment of the present application. The medical image tumor detection method device 40 of the embodiment of the present application includes:

[0060] Image acquisition module 41: used to acquire ultrasonic video images and contrast images of the detection site;

[0061] First image segmentation module 42: used to input the ultrasonic video image into the Mem-SAM2 model based on the double-layer memory bank mechanism. The Mem-SAM2 model based on the double-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasonic video image and generate segmentation labels for each ultrasonic video frame;

[0062] Second image segmentation module 43: used to superimpose the segmentation labels of each ultrasonic video frame onto the corresponding contrast image frame respectively, and segment each contrast image frame into a tumor region and a background region;

[0063] Feature extraction module 44: used to extract the temporal intensity curve of the tumor region and the background region to obtain the temporal features of the ultrasonic video image;

[0064] Tumor detection module 45: used to detect the benign and malignant nature of the tumor according to the temporal features.

[0065] It should be noted that for the information interaction, execution process, etc. between the above-mentioned devices / units, since they are based on the same concept as the method embodiment of the present application, their specific functions and the technical effects brought can be specifically referred to the method embodiment part, and will not be elaborated here.

[0066] The device provided in the embodiment of the present application can be applied to the foregoing method embodiment. For details, please refer to the description of the above method embodiment, and will not be elaborated here.

[0067] Please refer to Figure 4 , which is a schematic structural diagram of a computer device according to an embodiment of the present application. The computer device 50 includes:

[0068] A memory 51 storing executable program instructions;

[0069] A processor 52 connected to the memory 51;

[0070] The processor 52 is used to call the executable program instructions stored in the memory 51 and perform the following steps: obtaining the ultrasonic video image and the contrast image of the detection site; inputting the ultrasonic video image into the Mem-SAM2 model based on the double-layer memory bank mechanism, and the Mem-SAM2 model based on the double-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasonic video image to generate the segmentation label of each ultrasonic video frame; respectively superimposing the segmentation label of each ultrasonic video frame on the corresponding contrast image frame, and segmenting each contrast image frame into a tumor region and a background region; extracting the temporal intensity curve of the tumor region and the background region to obtain the temporal feature of the ultrasonic video image; performing tumor benign and malignant detection according to the temporal feature.

[0071] Among them, the processor 52 can also be called a CPU (Central Processing Unit, central processing unit). The processor 52 may be an integrated circuit chip with signal processing capabilities. The processor 52 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0072] Please refer to Figure 5, which is a schematic structural diagram of the storage medium according to an embodiment of the present application. The storage medium according to the embodiment of the present application stores program instructions 61 that can implement the following steps: obtaining an ultrasonic video image and a contrast image of a detection site; inputting the ultrasonic video image into the Mem-SAM2 model based on a double-layer memory bank mechanism, and the Mem-SAM2 model based on the double-layer memory bank mechanism uses a short-term memory bank and a long-term memory bank to segment the ultrasonic video image to generate segmentation labels for each ultrasonic video frame; respectively superimposing the segmentation labels of each ultrasonic video frame onto the corresponding contrast image frame, and segmenting each contrast image frame into a tumor region and a background region; extracting a temporal intensity curve for the tumor region and the background region to obtain the temporal features of the ultrasonic video image; and performing tumor benign and malignant detection based on the temporal features. Among them, the program instructions 61 can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network computer device, etc.) or a processor to execute all or part of the steps of the methods according to various embodiments of the present application. The foregoing storage medium includes: various media that can store program instructions such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc, or a terminal computer device such as a computer, a server, a mobile phone, or a tablet. Among them, the server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0073] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be an indirect coupling or communication connection through some interfaces, devices, or units, and can be in an electrical, mechanical, or other form.

[0074] In addition, each functional unit in various embodiments of the present application may be integrated into one processing unit, may exist separately as individual physical units, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units. The above is only the implementation mode of the present application, and does not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A medical imaging tumor detection method, characterized in that: include: Acquire ultrasonic video images and contrast images of the detection site; Inputting the ultrasound video image into a Mem-SAM2 model based on a double-layer memory bank mechanism, wherein the Mem-SAM2 model based on a double-layer memory bank mechanism segments the ultrasound video image using a short-term memory bank and a long-term memory bank to generate a segmentation label for each ultrasound video frame; Superimposing the segmentation label of each ultrasound video frame onto the corresponding contrast image frame, and segmenting each contrast image frame into a tumor area and a background area; Extracting time-series intensity curves of the tumor area and the background area to obtain time-series features of the ultrasound video image; The benign and malignant nature of the tumor is detected according to the time series characteristics.

2. The medical imaging tumor detection method according to claim 1, characterized in that: The step of obtaining the ultrasonic video image and the contrast imaging image of the detection part further includes: The radiomics features are extracted from the ultrasound video image; the radiomics features include organ size, shape, structure, metabolism, blood flow and / or intensity features.

3. The medical imaging tumor detection method according to claim 2, characterized in that: The step of inputting the ultrasonic video image into the Mem-SAM2 model based on the double-layer memory bank mechanism, and before the Mem-SAM2 model based on the double-layer memory bank mechanism uses the short-term memory bank and the long-term memory bank to segment the ultrasonic video image, further comprises: The ultrasound video image is preprocessed, and the preprocessing includes standardization and normalization processing.

4. The medical imaging tumor detection method according to claim 3, characterized in that: The Mem-SAM2 model based on the double-layer memory bank mechanism includes a short-term memory bank and a long-term memory bank. The short-term memory bank is used to store key information at the current time point, and the key information includes the current video frame and video frames similar to the current video frame. When the short-term memory bank is full, the video frame with the highest quality and the highest similarity to the current video frame is transferred to the long-term memory bank for storage, and the video frame with the lowest similarity to the current video frame is discarded, and a new video frame is stored.

5. The medical imaging tumor detection method according to claim 1, characterized in that: The extracting of the time series intensity curves of the tumor area and the background area to obtain the time series features of the ultrasound video image is specifically: The timing intensity curve is used to reflect the brightness changes of the tumor area and the background area on the time axis, analyze the wash-in and wash-out characteristics reflected by the timing intensity curve and the brightness difference between the tumor area and the background area during imaging, and extract the timing characteristics that can reflect the benign and malignant nature of the tumor from the ultrasound video image based on the brightness difference analysis results.

6. The medical imaging tumor detection method according to any one of claims 1 to 5, characterized in that: The detection of benign and malignant tumors according to the time series characteristics is specifically: The imaging omics features and the temporal features are input into a classification model together, and the classification model outputs a benign or malignant tumor classification result.

7. The medical imaging tumor detection method according to claim 6, characterized in that: The classification model uses a support vector machine or a random forest classifier.

8. A medical imaging tumor detection device, characterized in that: include: Image acquisition module: used to acquire ultrasonic video images and contrast images of the detection site; A first image segmentation module: used for inputting the ultrasound video image into the Mem-SAM2 model based on the double-layer memory bank mechanism, wherein the Mem-SAM2 model based on the double-layer memory bank mechanism segments the ultrasound video image using a short-term memory bank and a long-term memory bank to generate a segmentation label for each ultrasound video frame; A second image segmentation module: used for superimposing the segmentation label of each ultrasound video frame onto the corresponding contrast image frame, and segmenting each contrast image frame into a tumor area and a background area; Feature extraction module: used for extracting time series intensity curves of the tumor area and the background area to obtain time series features of the ultrasound video image; Tumor detection module: used to detect whether the tumor is benign or malignant based on the time series characteristics.

9. A computer device, characterized in that: The computer device includes a processor and a memory coupled to the processor, wherein: The memory stores program instructions for implementing the medical imaging tumor detection method according to any one of claims 1 to 7; The processor is used to execute the program instructions stored in the memory to control the medical imaging tumor detection method.

10. A storage medium, characterized in that: Program instructions executable by a processor are stored, and the program instructions are used to execute the medical imaging tumor detection method according to any one of claims 1 to 7.

Images