Lesion position identification method and system based on nuclear magnetism and CT image fusion

Abstract

The invention discloses a lesion position recognition method and system based on nuclear magnetism and CT image fusion, and belongs to the technical field of image recognition, and the method comprises the following steps: S1, image collection and preprocessing: collecting nuclear magnetism and CT real-time images of a patient, and preprocessing the nuclear magnetism and CT real-time images of the patient; s2, image registration and fusion: performing registration and fusion on the nuclear magnetism and CT real-time images of the patient to generate a nuclear magnetism and CT fusion image of the patient; s3, lesion position recognition: performing image recognition on the generated patient nuclear magnetism and CT fusion image according to a lesion position recognition model, and determining the lesion position of the patient. The problems that the lesion position of the patient cannot be effectively recognized in the prior art, so that the lesion position recognition precision and efficiency of the patient are low are solved. The lesion position of the patient can be effectively recognized based on nuclear magnetism and CT image fusion, the recognition precision and efficiency of the lesion position of the patient can be improved, and powerful support can be provided for clinical diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for identifying the position of a lesion based on the fusion of nuclear magnetic resonance and CT images. Background Art

[0002] The image modalities provided by medical imaging for clinical medical diagnosis are becoming increasingly diverse. Due to different imaging principles, multimodal medical images enable medical images of different modalities to have different characteristics and advantages. Therefore, the fusion of images with different characteristics and advantages can reduce the error rate of clinical diagnosis.

[0003] The Chinese patent with publication number CN115471432A discloses a method and system for fusing nuclear magnetic resonance images and CT images, wherein the method comprises: performing wavelet decomposition on the nuclear magnetic resonance image and the CT image and calculating the information gradient value of the decomposed image; constructing a fusion coefficient according to the information gradient value; calculating the information value of the nuclear magnetic resonance wavelet image and the CT wavelet image; constructing a fusion formula according to the information value; constructing an image fusion model according to the fusion formula and the fusion coefficient; fusing the nuclear magnetic resonance image and the CT image using the image fusion model to obtain a fused image; constructing an image fusion model according to the information gradient value and the information value of the image to be fused, and fusing the nuclear magnetic resonance image and the CT image based on the image fusion model, which can not only improve the clarity of the image, but also make up for the defect of insufficient information of a single nuclear magnetic resonance image or a CT image, greatly improve the doctor's interpretation of the image, and provide a favorable imaging basis for subsequent lesion positioning; however, the patent has the following defects: Existing technologies cannot effectively identify the location of patient lesions based on the fusion of MRI and CT images, resulting in low accuracy and efficiency in identifying the location of patient lesions and failing to provide strong support for clinical diagnosis. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for identifying the location of a lesion based on the fusion of MRI and CT images, which can effectively identify the location of a patient's lesion based on the fusion of MRI and CT images, improve the accuracy and efficiency of identifying the location of a patient's lesion, provide strong support for clinical diagnosis, and solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: The lesion location recognition method based on fusion of nuclear magnetic resonance and CT images includes the following steps: S1. Image acquisition and preprocessing: Based on the image acquisition terminal, the patient's nuclear magnetic resonance and CT real-time images are acquired, and the acquired patient's nuclear magnetic resonance and CT real-time images are preprocessed to remove the patient's nuclear magnetic resonance and CT real-time image noise, adjust the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images, and enhance the patient's nuclear magnetic resonance and CT real-time images; S2. Image registration and fusion: A transformation-based registration method is used to register the patient's nuclear magnetic resonance and CT real-time images, and a transform-domain-based image fusion method is used to generate a patient's nuclear magnetic resonance and CT fusion image containing the patient's nuclear magnetic resonance real-time image and the patient's CT real-time image; S3. Lesion location recognition: Based on machine learning technology, a lesion location recognition model based on the fusion of MRI and CT images is trained. Image recognition is performed on the generated patient MRI and CT fusion images according to the lesion location recognition model based on the fusion of MRI and CT images to determine the patient lesion location.

[0006] Preferably, in S1, acquiring the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal includes: Patient MRI image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the MRI to obtain the patient's real-time MRI image; Patient CT image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the CT machine to obtain the patient's real-time CT image; Wherein, based on the real-time nuclear magnetic resonance image and the real-time CT image of the patient, the real-time nuclear magnetic resonance image and CT image of the patient based on the image acquisition terminal are determined.

[0007] Preferably, in S1, preprocessing the collected patient nuclear magnetic resonance and CT real-time images includes: Image denoising: Based on filters, filter processing is performed on the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal to remove noise from the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal; Image brightness and contrast adjustment: Based on image segmentation technology, the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal are adjusted. The brightness of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal is kept consistent through interpolation or reconstruction technology. Based on the adaptive adjustment method of global contrast and adjacent pixel relationship, the pixel contrast is automatically adjusted to determine high-definition patient nuclear magnetic resonance and CT real-time images; Image enhancement: Based on image enhancement technology, high-definition patient MRI and real-time CT images are enhanced. The edges of high-definition patient MRI and real-time CT images are identified based on edge detection algorithms. The edges of patient MRI and real-time CT images are detected by calculating the gradient values ​​of neighborhood pixels and enhanced to highlight the edges in high-definition patient MRI and real-time CT images.

[0008] Preferably, in S1, preprocessing the collected patient nuclear magnetic resonance and CT real-time images includes: Extract the patient's MRI and CT real-time images after edge enhancement as the target image; Extracting grayscale values ​​corresponding to edge pixel blocks in the target image to form a first grayscale value set; Extracting grayscale values ​​corresponding to non-edge pixel blocks in the target image to form a second grayscale value set; Obtaining a grayscale relative coefficient using the first grayscale value set and the second grayscale value set; The grayscale relative coefficient is obtained by the following formula:

[0009] Wherein, W represents the grayscale relative coefficient; n represents the number of grayscale values ​​included in the first grayscale value set; m represents the number of grayscale values ​​included in the second grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; H z Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H mz represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set; Comparing the grayscale relative coefficient with a preset relative coefficient threshold; When the grayscale relative coefficient exceeds a preset relative coefficient threshold, no contrast adjustment is performed on the target image; When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the contrast of the target image is adjusted.

[0010] Preferably, when the grayscale relative coefficient does not exceed a preset relative coefficient threshold, contrast adjustment is performed on the target image, including: When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the grayscale value included in the second grayscale value set of the target image is retrieved; Obtaining a contrast adjustment coefficient using the grayscale values ​​included in the second grayscale value set; The contrast adjustment coefficient is obtained by the following formula:

[0011] Wherein, R represents the contrast adjustment coefficient; m represents the number of grayscale values ​​contained in the second grayscale value set; Hmi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; H z Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H np represents the grayscale average value corresponding to the n grayscale values ​​included in the first grayscale value set; W represents the grayscale relative coefficient; The contrast of the target image is adjusted by using the contrast adjustment coefficient, and the contrast-adjusted target image is obtained as an enhanced image that highlights the edge of the high-definition patient nuclear magnetic resonance and CT real-time images; The adjusted contrast value is obtained by the following formula:

[0012] Among them, D t represents the contrast value after adjustment; D represents the contrast value before adjustment; R represents the contrast adjustment coefficient; W represents the grayscale relative coefficient; n represents the number of grayscale values ​​contained in the first grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mz Represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set.

[0013] Preferably, in S2, registering and fusing the patient's nuclear magnetic resonance and CT real-time images includes: Image registration: Register the patient's nuclear magnetic resonance (MRI) and real-time CT images, and use a transformation-based registration method to make the patient's nuclear magnetic resonance (MRI) and real-time CT images correspond to each other in space; Image fusion: The registered patient MRI and CT real-time images are fused. The transform domain-based image fusion method is used to decompose the source image into sub-bands of different scales using wavelet transform, and the sub-bands are fused to generate a patient MRI and CT fused image that includes the patient MRI real-time image and the patient CT real-time image.

[0014] Preferably, in S3, based on machine learning technology, a lesion location recognition model based on fusion of nuclear magnetic resonance and CT images is trained, including: Collect data sources: Based on the need for lesion location identification based on MRI and CT image fusion, collect historical MRI and CT images of patients to determine the data source for model training; Data source division: The collected patient MRI and CT historical images are divided into training sets and test sets according to the ratio of 7:3; Model training: Based on the training set, the machine learning model is trained to determine the lesion location recognition model based on the fusion of MRI and CT images. Based on the test set, the performance of the trained lesion location recognition model based on the fusion of MRI and CT images is tested to determine whether the lesion location recognition model based on the fusion of MRI and CT images can achieve the expected effect and to determine the optimal lesion location recognition model based on the fusion of MRI and CT images.

[0015] Preferably, in S3, image recognition is performed on the generated patient nuclear magnetic resonance and CT fusion images according to a lesion location recognition model based on nuclear magnetic resonance and CT image fusion, including: Obtain the optimal lesion location recognition model based on fusion of MRI and CT images; Deploy the optimal lesion location recognition model based on fusion of MRI and CT images in the actual lesion location recognition environment; The generated patient MRI and CT fusion images are subjected to image recognition according to the optimal lesion location recognition model based on MRI and CT image fusion, and the patient lesion location is identified to determine the patient lesion location.

[0016] Preferably, in S3, after the patient's lesion location is identified, the patient's lesion location is marked and displayed in a visual form on the user interface.

[0017] According to another aspect of the present invention, a lesion location recognition system based on fusion of nuclear magnetic resonance and CT images is provided, which is used to implement the above-mentioned lesion location recognition method based on fusion of nuclear magnetic resonance and CT images, comprising: An image acquisition unit is used to acquire real-time nuclear magnetic resonance images and real-time CT images of patients, and determine real-time nuclear magnetic resonance and CT images of patients; Image preprocessing unit, used to perform denoising, brightness and contrast adjustment and image enhancement processing on the collected patient MRI and CT real-time images; An image fusion unit is used to fuse the patient's nuclear magnetic resonance and CT real-time images to generate a fusion image of the patient's nuclear magnetic resonance and CT; A lesion location recognition unit is used to perform image recognition on the generated patient MRI and CT fusion images and identify the patient's lesion location, thereby determining the patient's lesion location; The user interface display unit is used to display the patient's lesion location to the user in a visual form.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention collects patient nuclear magnetic resonance and CT real-time images based on an image acquisition terminal, and pre-processes the collected patient nuclear magnetic resonance and CT real-time images, removes noise from the patient nuclear magnetic resonance and CT real-time images, adjusts the brightness and contrast of the patient nuclear magnetic resonance and CT real-time images, and enhances the patient nuclear magnetic resonance and CT real-time images. The patient nuclear magnetic resonance and CT real-time images are registered based on a transformation-based registration method, and a transformation-domain-based image fusion method is used to generate a patient nuclear magnetic resonance and CT fusion image containing the patient nuclear magnetic resonance real-time image and the patient CT real-time image, so as to facilitate subsequent accurate recognition of the patient nuclear magnetic resonance and CT fusion image.

[0019] 2. The present invention is based on machine learning technology to train a lesion location recognition model based on the fusion of MRI and CT images. According to the lesion location recognition model based on the fusion of MRI and CT images, image recognition is performed on the generated patient MRI and CT fusion images and the patient lesion location is identified, the patient lesion location is determined, and the patient lesion location is marked, and the patient lesion location is displayed in a visual form on the user interface. The patient lesion location can be effectively identified based on the fusion of MRI and CT images, the patient lesion location can be improved, and the patient lesion location recognition accuracy and efficiency can be improved, which can provide strong support for clinical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the lesion location identification method based on fusion of nuclear magnetic resonance and CT images of the present invention; Figure 2 It is a process diagram of the lesion location identification method based on fusion of nuclear magnetic resonance and CT images of the present invention; Figure 3 It is a module diagram of the lesion location recognition system based on fusion of nuclear magnetic resonance and CT images of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] In order to solve the problem that the existing technology cannot effectively identify the location of patient lesions based on fusion of MRI and CT images, resulting in low accuracy and efficiency in identifying the location of patient lesions and failing to provide strong support for clinical diagnosis, please refer to Figure 1-Figure 3 , this embodiment provides the following technical solutions: Embodiment 1

[0023] The lesion location recognition method based on fusion of nuclear magnetic resonance and CT images includes the following steps: S1. Image acquisition and preprocessing: Based on the image acquisition terminal, the patient's nuclear magnetic resonance and CT real-time images are acquired, and the acquired patient's nuclear magnetic resonance and CT real-time images are preprocessed to remove the patient's nuclear magnetic resonance and CT real-time image noise, adjust the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images, and enhance the patient's nuclear magnetic resonance and CT real-time images; In this embodiment, as a preferred technical solution of the present invention, the real-time MRI and CT images of the patient are collected based on the image acquisition terminal, including: Patient MRI image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the MRI to obtain the patient's real-time MRI image; Patient CT image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the CT machine to obtain the patient's real-time CT image; Wherein, based on the real-time nuclear magnetic resonance image and the real-time CT image of the patient, the real-time nuclear magnetic resonance image and CT image of the patient based on the image acquisition terminal are determined.

[0024] It should be noted that nuclear magnetic resonance is a biological magnetic spin imaging technology, which utilizes the characteristics of atomic nucleus spin motion in an external magnetic field, generates signals after being stimulated by radio frequency pulses, detects with a detector and inputs into a computer, and displays images on the screen after processing and conversion; the amount of information provided by nuclear magnetic resonance is not only greater than many other imaging technologies in medical imaging, but also different from existing imaging technologies. Therefore, it has great potential advantages in the diagnosis of diseases, and can directly produce cross-sectional, sagittal, coronal and various oblique tomographic images, without the artifacts seen in CT detection, without the need for injection of contrast agents, without ionizing radiation, and without adverse effects on the body; therefore, the real-time monitoring and continuous acquisition of the patient's lesion site by a nuclear magnetic resonance imager can provide high-resolution soft tissue contrast by obtaining real-time nuclear magnetic images of the patient.

[0025] It should be noted that CT, or computer tomography, is a medical imaging technology that uses X-ray beams to perform layered scans of the human body and uses computer processing to produce detailed images of the body's internal structures. CT scanning is widely used in the diagnosis of various diseases due to its fast and clear imaging capabilities. CT images reflect the degree of X-ray absorption by organs and tissues in different grayscales, have high-density resolution, and can clearly display soft tissue and bone structures. Therefore, by using a CT machine to monitor and continuously collect the patient's lesion site in real time, and obtaining the patient's real-time CT image, detailed information on the bone and calcification structure can be provided.

[0026] In this embodiment, as a preferred technical solution of the present invention, the collected patient nuclear magnetic resonance and CT real-time images are preprocessed, including: Image denoising: Based on filters, filter processing is performed on the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal to remove noise from the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal; It should be noted that the filter is a filtering circuit composed of capacitors, inductors and resistors. The filter can effectively filter out a specific frequency point in the power line or frequencies other than the frequency point to obtain a power signal of a specific frequency, or eliminate a power signal after eliminating a specific frequency; therefore, the filter can be used to filter the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal, remove noise in the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal, and reduce interference.

[0027] Image brightness and contrast adjustment: Based on image segmentation technology, the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal are adjusted. The brightness of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal is kept consistent through interpolation or reconstruction technology. Based on the adaptive adjustment method of global contrast and adjacent pixel relationship, the pixel contrast is automatically adjusted to determine high-definition patient nuclear magnetic resonance and CT real-time images; Image enhancement: Based on image enhancement technology, high-definition patient MRI and real-time CT images are enhanced. The edges of high-definition patient MRI and real-time CT images are identified based on edge detection algorithms. The edges of patient MRI and real-time CT images are detected by calculating the gradient values ​​of neighborhood pixels and enhanced to highlight the edges in high-definition patient MRI and real-time CT images.

[0028] Specifically, in S1, the collected real-time MRI and CT images of the patient are preprocessed, including: Extract the patient's MRI and CT real-time images after edge enhancement as the target image; Extracting grayscale values ​​corresponding to edge pixel blocks in the target image to form a first grayscale value set; Extracting grayscale values ​​corresponding to non-edge pixel blocks in the target image to form a second grayscale value set; Obtaining a grayscale relative coefficient using the first grayscale value set and the second grayscale value set; The grayscale relative coefficient is obtained by the following formula:

[0029] Wherein, W represents the grayscale relative coefficient; n represents the number of grayscale values ​​included in the first grayscale value set; m represents the number of grayscale values ​​included in the second grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; Hz Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H mz represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set; Comparing the grayscale relative coefficient with a preset relative coefficient threshold; When the grayscale relative coefficient exceeds a preset relative coefficient threshold, no contrast adjustment is performed on the target image; When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the contrast of the target image is adjusted.

[0030] The technical effect of the above technical solution is: by extracting the edge-enhanced patient nuclear magnetic resonance and CT real-time images as the target image, the technical solution can make full use of the image clarity improvement brought by the edge enhancement technology. Edge enhancement helps to highlight the key structures or lesion areas in the image, providing more valuable information for subsequent analysis and processing. The technical solution extracts the grayscale values ​​corresponding to the edge pixel blocks and non-edge pixel blocks in the target image respectively to form two grayscale value sets. This step realizes the refined analysis of the grayscale characteristics of the image. Grayscale value is the basic data in image analysis and is of great significance for understanding the texture, contrast and other characteristics of the image. By introducing the grayscale relative coefficient, the technical solution can comprehensively consider the grayscale distribution characteristics of the edge area and the non-edge area, as well as the overall grayscale level of the target image. The calculation of this coefficient not only considers the number and specific values ​​of the grayscale values, but also introduces parameters such as the central grayscale value and the grayscale median value, making the evaluation result more comprehensive and accurate. According to the comparison result of the grayscale relative coefficient and the preset relative coefficient threshold, the technical solution can intelligently decide whether to adjust the contrast of the target image. This processing method based on automatic adjustment of image characteristics avoids the subjectivity and uncertainty of human intervention and improves the efficiency and accuracy of processing. Since the technical solution is based on objective analysis of the grayscale value of the image to adjust the contrast, the processing effect is relatively stable and will not fluctuate due to differences in personal experience or subjective judgment. This is of great significance for maintaining the consistency and reliability of medical image processing. This technical solution is not only suitable for the processing of patient nuclear magnetic resonance and CT real-time images, but can also be extended to other types of medical image processing. At the same time, with the continuous development of image processing technology, the relevant parameters and algorithms in the technical solution can also be further optimized and improved to adapt to a wider and more complex application scenarios.

[0031] In summary, the technical effects of the above technical solutions in terms of performance indicators are mainly reflected in the evaluation and utilization of edge enhancement effects, refinement of gray value analysis, innovative application of gray relative coefficients, intelligent contrast adjustment, stability of processing effects, and adaptability and scalability. These technical effects jointly improve the efficiency and accuracy of medical image processing and provide doctors with more reliable and valuable diagnostic basis.

[0032] Specifically, when the grayscale relative coefficient does not exceed a preset relative coefficient threshold, contrast adjustment is performed on the target image, including: When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the grayscale value included in the second grayscale value set of the target image is retrieved; Obtaining a contrast adjustment coefficient using the grayscale values ​​included in the second grayscale value set; The contrast adjustment coefficient is obtained by the following formula:

[0033] Wherein, R represents the contrast adjustment coefficient; m represents the number of grayscale values ​​contained in the second grayscale value set; H mi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; H z Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H np represents the grayscale average value corresponding to the n grayscale values ​​included in the first grayscale value set; W represents the grayscale relative coefficient; The contrast of the target image is adjusted by using the contrast adjustment coefficient, and the contrast-adjusted target image is obtained as an enhanced image that highlights the edge of the high-definition patient nuclear magnetic resonance and CT real-time images; The adjusted contrast value is obtained by the following formula:

[0034] Among them, D t represents the contrast value after adjustment; D represents the contrast value before adjustment; R represents the contrast adjustment coefficient; W represents the grayscale relative coefficient; n represents the number of grayscale values ​​contained in the first grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mz Represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set.

[0035] The technical effect of the above technical solution is: by introducing the contrast adjustment coefficient R, the coefficient integrates multiple factors of the second grayscale value set (i.e., the grayscale value of the non-edge area), including the number of grayscale values, the specific grayscale value, the central grayscale value Hz of the target image, the grayscale median value Hnz and the grayscale average value Hnp of the first grayscale value set, and the grayscale relative coefficient W. This comprehensive consideration makes the contrast adjustment more accurate and can be personalized for the characteristics of different images. The contrast adjustment coefficient R in the technical solution and the contrast value Dt after the target image is adjusted are dynamically calculated based on multiple factors, which means that for different target images, the degree and method of contrast adjustment will be different. This dynamic adjustment capability enables the technical solution to adapt to more diverse image processing needs. Through contrast adjustment, especially when the grayscale relative coefficient does not exceed the preset threshold, the technical solution can significantly enhance the edges in the target image, making the edges clearer and more prominent. This is particularly important for medical images, because edge information often contains important clues to lesion areas or key structures. The above technical solution calculates the contrast adjustment coefficient and the adjusted contrast value by comprehensively considering multiple factors, which makes it robust and adaptable to different types of medical images (such as MRI and CT images). Regardless of how the background, brightness or contrast of the image changes, the technical solution can effectively process it and obtain a satisfactory edge highlighting effect. Ultimately, through contrast adjustment and edge highlighting, the technical solution can significantly improve the clarity and recognition of medical images, thereby providing doctors with more accurate and reliable diagnostic basis. This is of great significance to improving medical standards and protecting patient health.

[0036] In summary, the technical effects of this technical solution in terms of performance indicators are mainly reflected in the accuracy of contrast adjustment, dynamic adjustment capability, edge highlighting effect, computational efficiency, robustness and adaptability. These effects jointly improve the processing quality and diagnostic accuracy of medical images.

[0037] S2. Image registration and fusion: A transformation-based registration method is used to register the patient's nuclear magnetic resonance and CT real-time images, and a transform-domain-based image fusion method is used to generate a patient's nuclear magnetic resonance and CT fusion image containing the patient's nuclear magnetic resonance real-time image and the patient's CT real-time image; In this embodiment, as a preferred technical solution of the present invention, the patient's nuclear magnetic resonance and CT real-time images are registered and fused, including: Image registration: Register the patient's nuclear magnetic resonance (MRI) and real-time CT images, and use a transformation-based registration method to make the patient's nuclear magnetic resonance (MRI) and real-time CT images correspond to each other in space; Image fusion: The registered patient MRI and CT real-time images are fused. The transform domain-based image fusion method is used to decompose the source image into sub-bands of different scales using wavelet transform. The sub-bands are then fused to generate a patient MRI and CT fused image containing the patient MRI real-time image and the patient CT real-time image, so as to retain the advantageous information of the patient MRI real-time image and the patient CT real-time image.

[0038] S3. Lesion location recognition: Based on machine learning technology, a lesion location recognition model based on the fusion of MRI and CT images is trained. Image recognition is performed on the generated patient MRI and CT fusion images according to the lesion location recognition model based on the fusion of MRI and CT images to determine the patient lesion location.

[0039] In this embodiment, as a preferred technical solution of the present invention, based on machine learning technology, a lesion location recognition model based on fusion of nuclear magnetic resonance and CT images is trained, including: Collect data sources: Based on the need for lesion location identification based on MRI and CT image fusion, collect historical MRI and CT images of patients to determine the data source for model training; Data source division: The collected patient MRI and CT historical images are divided into training sets and test sets according to the ratio of 7:3; Model training: Based on the training set, the machine learning model is trained to determine the lesion location recognition model based on the fusion of MRI and CT images. Based on the test set, the performance of the trained lesion location recognition model based on the fusion of MRI and CT images is tested to determine whether the lesion location recognition model based on the fusion of MRI and CT images can achieve the expected effect and to determine the optimal lesion location recognition model based on the fusion of MRI and CT images.

[0040] In this embodiment, as a preferred technical solution of the present invention, image recognition is performed on the generated patient nuclear magnetic resonance and CT fusion images according to a lesion location recognition model based on nuclear magnetic resonance and CT image fusion, including: Obtain the optimal lesion location recognition model based on fusion of MRI and CT images; Deploy the optimal lesion location recognition model based on fusion of MRI and CT images in the actual lesion location recognition environment; The generated patient MRI and CT fusion images are subjected to image recognition according to the optimal lesion location recognition model based on MRI and CT image fusion, and the patient lesion location is identified to determine the patient lesion location.

[0041] In this embodiment, as a preferred technical solution of the present invention, after the patient's lesion position is identified, the patient's lesion position is marked and displayed on the user interface in a visual form.

[0042] Embodiment 2 In order to better demonstrate the principle of lesion location recognition based on fusion of nuclear magnetic resonance and CT images, this embodiment now provides a lesion location recognition system based on fusion of nuclear magnetic resonance and CT images, which is used to implement the above-mentioned lesion location recognition method based on fusion of nuclear magnetic resonance and CT images, including: An image acquisition unit is used to acquire real-time nuclear magnetic resonance images and real-time CT images of patients, and determine real-time nuclear magnetic resonance and CT images of patients; Image preprocessing unit, used to perform denoising, brightness and contrast adjustment and image enhancement processing on the collected patient MRI and CT real-time images; An image fusion unit is used to fuse the patient's nuclear magnetic resonance and CT real-time images to generate a fusion image of the patient's nuclear magnetic resonance and CT; A lesion location recognition unit is used to perform image recognition on the generated patient MRI and CT fusion images and identify the patient's lesion location, thereby determining the patient's lesion location; The user interface display unit is used to display the patient's lesion location to the user in a visual form.

[0043] It should be noted that the patient's nuclear MRI and CT real-time images are acquired by the image acquisition unit, the acquired patient's nuclear MRI and CT real-time images are preprocessed by the image preprocessing unit, the patient's nuclear MRI and CT real-time images are noise-removed, the brightness and contrast of the patient's nuclear MRI and CT real-time images are adjusted, and the patient's nuclear MRI and CT real-time images are enhanced, the patient's nuclear MRI and CT real-time images are registered by the image fusion unit, and a transform domain-based image fusion method is used to generate a patient's nuclear MRI and CT fused image containing the patient's nuclear MRI real-time image and the patient's CT real-time image, the lesion position recognition unit performs image recognition on the generated patient's nuclear MRI and CT fused image to determine the patient's lesion position, and the user interface display unit displays the patient's lesion position to the user in a visual form, the patient's lesion position can be effectively identified based on the fusion of nuclear MRI and CT images, the patient's lesion position recognition accuracy and efficiency can be improved, and strong support can be provided for clinical diagnosis.

[0044] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0045] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A lesion location recognition method based on fusion of MRI and CT images, characterized in that: The steps include: S1. Image acquisition and preprocessing: Based on the image acquisition terminal, the patient's nuclear magnetic resonance and CT real-time images are acquired, and the acquired patient's nuclear magnetic resonance and CT real-time images are preprocessed to remove the patient's nuclear magnetic resonance and CT real-time image noise, adjust the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images, and enhance the patient's nuclear magnetic resonance and CT real-time images; S2. Image registration and fusion: A transformation-based registration method is used to register the patient's nuclear magnetic resonance and CT real-time images, and a transform-domain-based image fusion method is used to generate a patient's nuclear magnetic resonance and CT fusion image containing the patient's nuclear magnetic resonance real-time image and the patient's CT real-time image; S3. Lesion location recognition: Based on machine learning technology, a lesion location recognition model based on the fusion of MRI and CT images is trained. Image recognition is performed on the generated patient MRI and CT fusion images according to the lesion location recognition model based on the fusion of MRI and CT images to determine the patient lesion location.

2. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 1, characterized in that: In S1, the real-time MRI and CT images of the patient are collected based on the image acquisition terminal, including: Patient MRI image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the MRI to obtain the patient's real-time MRI image; Patient CT image acquisition: Real-time monitoring and continuous acquisition of the patient's lesion site based on the CT machine to obtain the patient's real-time CT image; Wherein, based on the real-time nuclear magnetic resonance image and the real-time CT image of the patient, the real-time nuclear magnetic resonance image and CT image of the patient based on the image acquisition terminal are determined.

3. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 2, characterized in that: In S1, the collected real-time MRI and CT images of the patient are preprocessed, including: Image denoising: Based on filters, filter processing is performed on the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal to remove noise from the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal; Image brightness and contrast adjustment: Based on image segmentation technology, the brightness and contrast of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal are adjusted. The brightness of the patient's nuclear magnetic resonance and CT real-time images based on the image acquisition terminal is kept consistent through interpolation or reconstruction technology. Based on the adaptive adjustment method of global contrast and adjacent pixel relationship, the pixel contrast is automatically adjusted to determine high-definition patient nuclear magnetic resonance and CT real-time images; Image enhancement: Based on image enhancement technology, high-definition patient MRI and real-time CT images are enhanced. The edges of high-definition patient MRI and real-time CT images are identified based on edge detection algorithms. The edges of patient MRI and real-time CT images are detected by calculating the gradient values ​​of neighborhood pixels and enhanced to highlight the edges in high-definition patient MRI and real-time CT images.

4. The method for identifying the position of a lesion based on fusion of nuclear magnetic resonance and CT images as claimed in claim 3, characterized in that: In S1, the collected real-time MRI and CT images of the patient are preprocessed, including: Extract the patient's MRI and CT real-time images after edge enhancement as the target image; Extracting grayscale values ​​corresponding to edge pixel blocks in the target image to form a first grayscale value set; Extracting grayscale values ​​corresponding to non-edge pixel blocks in the target image to form a second grayscale value set; Obtaining a grayscale relative coefficient using the first grayscale value set and the second grayscale value set; The grayscale relative coefficient is obtained by the following formula: Wherein, W represents the grayscale relative coefficient; n represents the number of grayscale values ​​included in the first grayscale value set; m represents the number of grayscale values ​​included in the second grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; H z Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H mz represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set; Comparing the grayscale relative coefficient with a preset relative coefficient threshold; When the grayscale relative coefficient exceeds a preset relative coefficient threshold, no contrast adjustment is performed on the target image; When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the contrast of the target image is adjusted.

5. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 4, characterized in that: When the grayscale relative coefficient does not exceed the preset relative coefficient threshold, the contrast of the target image is adjusted, including: When the grayscale relative coefficient does not exceed a preset relative coefficient threshold, the grayscale value included in the second grayscale value set of the target image is retrieved; Obtaining a contrast adjustment coefficient using the grayscale values ​​included in the second grayscale value set; The contrast adjustment coefficient is obtained by the following formula: Wherein, R represents the contrast adjustment coefficient; m represents the number of grayscale values ​​contained in the second grayscale value set; H mi represents the corresponding value of the i-th grayscale value contained in the second grayscale value set; H z Represents the central gray value of the target image; H nz represents the grayscale intermediate value corresponding to the n grayscale values ​​included in the first grayscale value set; H np represents the grayscale average value corresponding to the n grayscale values ​​included in the first grayscale value set; W represents the grayscale relative coefficient; The contrast of the target image is adjusted by using the contrast adjustment coefficient, and the contrast-adjusted target image is obtained as an enhanced image that highlights the edge of the high-definition patient nuclear magnetic resonance and CT real-time images; The adjusted contrast value is obtained by the following formula: Among them, D t represents the contrast value after adjustment; D represents the contrast value before adjustment; R represents the contrast adjustment coefficient; W represents the grayscale relative coefficient; n represents the number of grayscale values ​​contained in the first grayscale value set; H ni represents the corresponding value of the i-th grayscale value contained in the first grayscale value set; H mz Represents the grayscale intermediate value corresponding to the m grayscale values ​​included in the second grayscale value set.

6. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 3, characterized in that: In S2, the patient's nuclear magnetic resonance and CT real-time images are registered and fused, including: Image registration: Register the patient's nuclear magnetic resonance (MRI) and real-time CT images, and use a transformation-based registration method to make the patient's nuclear magnetic resonance (MRI) and real-time CT images correspond to each other in space; Image fusion: The registered patient MRI and CT real-time images are fused. The image fusion method based on transform domain is adopted. The source image is decomposed into sub-bands of different scales by wavelet transform. The sub-bands are then fused to generate a patient MRI and CT fused image containing the patient MRI real-time image and the patient CT real-time image.

7. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 6, characterized in that: In S3, based on machine learning technology, a lesion location recognition model based on fusion of nuclear magnetic resonance and CT images is trained, including: Collect data sources: Based on the need for lesion location identification based on MRI and CT image fusion, collect historical MRI and CT images of patients to determine the data source for model training; Data source division: The collected patient MRI and CT historical images are divided into training sets and test sets according to the ratio of 7:3; Model training: Based on the training set, the machine learning model is trained to determine the lesion location recognition model based on the fusion of MRI and CT images. Based on the test set, the performance of the trained lesion location recognition model based on the fusion of MRI and CT images is tested to determine whether the lesion location recognition model based on the fusion of MRI and CT images can achieve the expected effect and to determine the optimal lesion location recognition model based on the fusion of MRI and CT images.

8. The method for identifying lesion positions based on fusion of nuclear magnetic resonance and CT images as claimed in claim 7, characterized in that: In S3, image recognition is performed on the generated fusion image of the patient's MRI and CT images according to a lesion location recognition model based on fusion of MRI and CT images, including: Obtain the optimal lesion location recognition model based on fusion of MRI and CT images; Deploy the optimal lesion location recognition model based on fusion of MRI and CT images in the actual lesion location recognition environment; The generated patient MRI and CT fusion images are subjected to image recognition according to the optimal lesion location recognition model based on MRI and CT image fusion, and the patient lesion location is identified to determine the patient lesion location.

9. The method for identifying lesion location based on fusion of nuclear magnetic resonance and CT images as claimed in claim 8, characterized in that: In S3, after the patient's lesion position is identified, the patient's lesion position is marked and displayed on the user interface in a visual form.

10. A lesion location recognition system based on fusion of nuclear magnetic resonance and CT images, used to implement the lesion location recognition method based on fusion of nuclear magnetic resonance and CT images as claimed in claim 7, characterized in that: include: An image acquisition unit is used to acquire real-time nuclear magnetic resonance images and real-time CT images of patients, and determine real-time nuclear magnetic resonance and CT images of patients; Image preprocessing unit, used to perform denoising, brightness and contrast adjustment and image enhancement processing on the collected patient MRI and CT real-time images; An image fusion unit is used to fuse the patient's nuclear magnetic resonance and CT real-time images to generate a fusion image of the patient's nuclear magnetic resonance and CT; A lesion location recognition unit is used to perform image recognition on the generated patient MRI and CT fusion images and identify the patient's lesion location, thereby determining the patient's lesion location; The user interface display unit is used to display the patient's lesion location to the user in a visual form.

Images