A magnetic resonance-based lesion identification system and method

Abstract

The application provides a lesion recognition system and method based on magnetic resonance, and relates to the technical field of lesion recognition, and comprises the following steps: collecting a coil to scan a to-be-scanned part of a current patient to obtain a magnetic resonance imaging signal when the current patient enters a scanning area; performing image reconstruction on the magnetic resonance imaging signal after electromagnetic interference elimination to obtain a magnetic resonance image; inputting the magnetic resonance image into a pre-trained first lesion recognition model, and segmenting the magnetic resonance image into a plurality of image blocks and generating a first image block set and a second image block set according to a recognition result; subsequently, inputting the second image block set into the first lesion recognition model and adjusting the first image block set and the second image block set according to a recognition result; subsequently, inputting the first image block set into a pre-trained second lesion recognition model to perform lesion identification, and then performing image splicing to obtain a lesion identification magnetic resonance image. The beneficial effect is to speed up the lesion recognition stage and save the overall process time.

Description

Technical Field

[0001] This invention relates to the field of lesion identification technology, and in particular to a lesion identification system and method based on magnetic resonance imaging. Background Technology

[0002] Magnetic resonance imaging (MRI) offers numerous advantages in the medical field, including radiation-free imaging, high-contrast imaging, and high accuracy. Compared to other medical imaging techniques, MRI provides excellent soft tissue contrast, extremely high imaging resolution, and the ability to image in any plane, displaying detailed anatomical structures, blood flow, and metabolic information. These characteristics enable doctors to diagnose conditions more accurately. Therefore, MRI technology is widely used in the medical field and plays a crucial role in the diagnosis and treatment of diseases.

[0003] The relatively long scan time of MRI significantly limits its application in some real-time scenarios, such as surgical navigation and emergency care. Existing technologies, such as parallel MRI, can greatly improve scan speed. Therefore, it is necessary to optimize the speed of lesion identification after MRI image reconstruction to achieve rapid lesion identification from the patient's MRI images. Summary of the Invention

[0004] To address the problems existing in the prior art, the present invention provides a lesion identification system based on magnetic resonance imaging, comprising:

[0005] The image reconstruction module is used to control the acquisition coil to scan the area to be scanned of the current patient when the current patient enters the scanning area to obtain magnetic resonance imaging signal, and to perform image reconstruction after eliminating electromagnetic interference of the magnetic resonance imaging signal to obtain magnetic resonance image.

[0006] The lesion identification module, connected to the image reconstruction module, is used to input a pre-trained first lesion identification model into the magnetic resonance image, and to segment the magnetic resonance image into multiple image blocks according to the identification results, generating a first image block set and a second image block set. Then, the second image block set is input into the first lesion identification model, and the first and second image block sets are adjusted according to the identification results. Subsequently, the first image block set is input into the pre-trained second lesion identification model for lesion identification to obtain multiple corresponding labeled image blocks. The image blocks and labeled image blocks in the first image block set are stitched together to obtain a lesion-labeled magnetic resonance image.

[0007] Preferably, an interference receiving coil is also provided around the scanning area, and the image reconstruction module includes:

[0008] An interference cancellation unit is used to control the interference receiving coil to collect electromagnetic interference signals from interference sources in the space to the scanning area, and then input the electromagnetic interference signals and the magnetic resonance imaging signals into a pre-trained interference cancellation model to obtain the corresponding cancellation signal as the magnetic resonance imaging signal.

[0009] An image reconstruction unit, connected to the interference cancellation unit, is used to perform image reconstruction based on the magnetic resonance imaging signal using an image reconstruction algorithm to obtain the magnetic resonance image.

[0010] Preferably, the image reconstruction module further includes:

[0011] An image denoising unit, connected to the image reconstruction unit, is used to denoise the magnetic resonance image and use it as the magnetic resonance image;

[0012] The artifact correction unit is connected to the image denoising unit and is used to perform artifact correction processing on the magnetic resonance image to obtain the magnetic resonance image.

[0013] An image enhancement unit, connected to the artifact correction unit, is used to perform image enhancement processing on the magnetic resonance image and then use it as the magnetic resonance image.

[0014] Preferably, the lesion identification module includes:

[0015] The first recognition unit is used to input the magnetic resonance image into a pre-trained first recognition model, segment the magnetic resonance image into multiple image blocks containing lesions and multiple image blocks not containing lesions according to the recognition results, then select each image block containing lesions to generate the first image set, and generate a second image set according to the remaining image blocks;

[0016] The second recognition unit, connected to the first recognition unit, is used to input each of the image blocks in the second image set into the first recognition model, and to filter out the image blocks containing lesions from the second image block set according to the recognition results and add them to the first image block set.

[0017] The lesion identification unit, connected to the second identification unit, is used to input the first image block set into the second identification module to identify the boundaries of the lesions contained in each image block in the first image block set to obtain the corresponding identification image block. Then, the image blocks and the image blocks in the first image block set are stitched together to obtain the lesion identification magnetic resonance image.

[0018] This invention also provides a lesion identification method based on magnetic resonance imaging, applied to the aforementioned lesion identification system, wherein the lesion identification method includes:

[0019] Step S1: When the current patient enters the scanning area, the lesion identification system controls the acquisition coil to scan the area to be scanned of the current patient to obtain a magnetic resonance imaging signal. After electromagnetic interference is eliminated from the magnetic resonance imaging signal, the image is reconstructed to obtain a magnetic resonance image.

[0020] In step S2, the lesion identification system inputs a pre-trained first lesion identification model into the magnetic resonance image, and segments the magnetic resonance image into multiple image blocks according to the identification results to generate a first image block set and a second image block set. Then, the second image block set is input into the first lesion identification model, and the first image block set and the second image block set are adjusted according to the identification results. Then, the first image block set is input into the pre-trained second lesion identification model to identify lesions and obtain multiple corresponding labeled image blocks. The image blocks and labeled image blocks in the first image block set are stitched together to obtain a lesion-labeled magnetic resonance image.

[0021] Preferably, an interference receiving coil is also provided around the scanning area, and step S1 includes:

[0022] Step S11: The lesion identification system controls the acquisition coil to scan the area to be scanned of the current patient when the current patient enters the scanning area to obtain the magnetic resonance imaging signal;

[0023] Step S12: The lesion identification system controls the interference receiving coil to collect electromagnetic interference signals from interference sources in the space to the scanning area, and then inputs the electromagnetic interference signals and the magnetic resonance imaging signals into a pre-trained interference cancellation model to obtain the corresponding cancellation signals as the magnetic resonance imaging signals.

[0024] Step S13: The lesion identification system uses an image reconstruction algorithm to reconstruct the magnetic resonance image based on the magnetic resonance imaging signal.

[0025] Preferably, step S1 further includes:

[0026] Step S14: The lesion identification system performs image denoising on the magnetic resonance image and then uses it as the magnetic resonance image;

[0027] Step S15: The lesion identification system performs artifact correction processing on the magnetic resonance image and then uses it as the magnetic resonance image.

[0028] Step S16: The lesion identification system performs image enhancement processing on the magnetic resonance image and uses it as the magnetic resonance image.

[0029] Preferably, step S2 includes:

[0030] Step S21: The lesion recognition system inputs the magnetic resonance image into a pre-trained first recognition model, and divides the magnetic resonance image into multiple image blocks containing lesions and multiple image blocks not containing lesions according to the recognition results. Then, it selects each image block containing lesions to generate the first image set, and generates a second image set according to the remaining image blocks.

[0031] Step S22: The lesion recognition system inputs each of the image blocks in the second image set into the first recognition model, and selects the image blocks containing lesions from the second image block set according to the recognition results and adds them to the first image block set.

[0032] In step S23, the lesion identification system inputs the first image block set into the second identification module to identify the boundaries of the lesions contained in each image block in the first image block set to obtain the corresponding identified image block. Then, the system stitches the images of each image block and each image block in the first image block set together to obtain the lesion identification magnetic resonance image.

[0033] The above technical solution has the following advantages or beneficial effects: In the image reconstruction stage, electromagnetic interference is eliminated from the acquired magnetic resonance signal, reducing environmental electromagnetic interference and improving imaging quality; In the lesion identification stage, a first identification model with low scanning accuracy is first used to segment the part of the magnetic resonance image containing the lesion, and then a second identification model with high scanning accuracy is used to accurately identify the lesion boundary. This can ensure the accuracy of lesion boundary identification while speeding up the lesion identification process and saving the overall process time from scanning to lesion identification. Attached Figure Description

[0034] Figure 1 A schematic diagram of a lesion identification system based on magnetic resonance imaging is provided in a preferred embodiment of the present invention.

[0035] Figure 2 A flowchart illustrating a lesion identification method based on magnetic resonance imaging is provided as a preferred embodiment of the present invention.

[0036] Figure 3 This is a schematic diagram of a sub-process of step S1 in a preferred embodiment of the present invention.

[0037] Figure 4 This is a schematic diagram of the sub-process of step S2 in a preferred embodiment of the present invention. Detailed Implementation

[0038] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0039] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, a lesion identification system based on magnetic resonance imaging is provided, such as... Figure 1 As shown, it includes:

[0040] Image reconstruction module 1 is used to control the acquisition coil to scan the area to be scanned of the current patient when the current patient enters the scanning area to obtain magnetic resonance imaging signal, and to perform image reconstruction after eliminating electromagnetic interference of the magnetic resonance imaging signal to obtain magnetic resonance image.

[0041] The lesion identification module 2, connected to the image reconstruction module 1, is used to input the magnetic resonance image into a pre-trained first lesion identification model, and to segment the magnetic resonance image into multiple image blocks according to the identification results, generating a first image block set and a second image block set. Then, the second image block set is input into the first lesion identification model, and the first and second image block sets are adjusted according to the identification results. Then, the first image block set is input into the pre-trained second lesion identification model to identify the lesion and obtain multiple corresponding labeled image blocks. The image blocks and labeled image blocks in the first image block set are stitched together to obtain the lesion-labeled magnetic resonance image.

[0042] In a preferred embodiment of the present invention, an interference receiving coil is further provided around the scanning area, such as... Figure 1 As shown, image reconstruction module 1 includes:

[0043] The signal acquisition module 11 is used to control the acquisition coil to scan the area to be scanned of the current patient when the current patient enters the scanning area to obtain magnetic resonance imaging signals;

[0044] Interference cancellation unit 12 is connected to signal acquisition module 11 and is used to control interference receiving coil to acquire electromagnetic interference signals from interference sources in the space to the scanning area. Then, the electromagnetic interference signal and magnetic resonance imaging signal are input into the pre-trained interference cancellation model to obtain the corresponding cancellation signal as the magnetic resonance imaging signal.

[0045] The image reconstruction unit 13 is connected to the interference cancellation unit 12 and is used to perform image reconstruction based on the magnetic resonance imaging signal using an image reconstruction algorithm to obtain a magnetic resonance image.

[0046] Specifically, in this embodiment, in order to avoid the interference of environmental sources on the scanned magnetic resonance imaging signal, which would lead to problems such as blurriness, unclear boundaries, and excessive noise in the later obtained magnetic resonance image, interference receiving coils are also set up around the scanning area to receive electromagnetic interference signals from the environment. Then, the collected magnetic resonance imaging signal and the received electromagnetic interference signal are input into a pre-trained electromagnetic interference cancellation model to perform electromagnetic interference cancellation and obtain the corresponding cancellation signal as the magnetic resonance imaging signal. Subsequently, an image reconstruction algorithm is used to reconstruct the image based on the magnetic resonance imaging signal to obtain the corresponding magnetic resonance image.

[0047] In a preferred embodiment of the present invention, such as Figure 1 As shown, image reconstruction module 2 also includes:

[0048] The image denoising unit 14 is connected to the image reconstruction unit 13 and is used to denoise the magnetic resonance image and then use it as the magnetic resonance image.

[0049] The artifact correction unit 15 is connected to the image denoising unit 14 and is used to perform artifact correction processing on the magnetic resonance image and then use it as the magnetic resonance image.

[0050] The image enhancement unit 16 is connected to the artifact correction unit 15 and is used to perform image enhancement processing on the magnetic resonance image to obtain the magnetic resonance image.

[0051] Specifically, in this embodiment, after the corresponding magnetic resonance image is obtained through image reconstruction, image optimization is required, including image denoising, artifact correction, and image enhancement.

[0052] Noise removal processes include:

[0053] Statistical filters: Statistical filters are methods based on the statistical characteristics of image pixels, including mean filters, median filters, etc. These filters remove noise by calculating the mean or median of pixels within a window surrounding a given pixel.

[0054] Methods based on partial differential equations: Commonly used methods include diffusion filtering and total variational regularization. These methods smooth images and remove noise by utilizing partial differential equation models.

[0055] Wavelet transform: Wavelet transform decomposes an image into sub-bands of different frequencies, removes noise from the high-frequency sub-bands, and finally reconstructs the image. Wavelet transform can effectively remove Gaussian noise and salt-and-pepper noise.

[0056] Non-Local Means (NLmeans) Filtering: NLmeans filtering is a method based on image patch similarity. It estimates noise by comparing the similarity between different pixels and removes noise using a weighted average of similar pixels.

[0057] Dictionary-based learning methods: These methods learn a dictionary and represent image patches as sparse linear combinations of the dictionary's elements. By controlling the sparsity coefficients, noise can be removed while preserving image details.

[0058] Select the above methods for noise reduction based on actual needs;

[0059] Artifact correction processing includes: gradient nonlinear correction, frequency correction, and artifact correction by using motion correction techniques, such as phase dephase correction and navigation techniques, to reduce motion-induced artifacts.

[0060] Image enhancement processing includes methods such as contrast enhancement, sharpening enhancement, and local enhancement.

[0061] Image restoration processing includes: artifact removal: artifacts are removed by using physical model-based reconstruction algorithms, such as Fast Fourier Transform (FFT) and slice correction; spatial domain reconstruction: image restoration is achieved by applying and optimizing regularization constraints in the sampling domain, such as total variation regularization and sample sparsity.

[0062] In a preferred embodiment of the present invention, such as Figure 1 As shown, the lesion identification module 2 includes:

[0063] The first recognition unit 21 is used to input the magnetic resonance image into a pre-trained first recognition model, and to segment the magnetic resonance image into multiple image blocks containing lesions and multiple image blocks not containing lesions according to the recognition results. Then, the image blocks containing lesions are selected to generate a first image set, and a second image set is generated according to the remaining image blocks.

[0064] The second recognition unit 22 is connected to the first recognition unit 21 and is used to input each image block in the second image set into the first recognition model, and to filter out the image blocks containing lesions from the second image block set according to the recognition results and add them to the first image block set.

[0065] The lesion identification unit 23 is connected to the second identification unit 22 and is used to input the first image block set into the second identification module to identify the boundaries of the lesions contained in each image block in the first image block set to obtain the corresponding identification image block. Then, the image blocks and the image blocks in the first image block set are stitched together to obtain the lesion identification magnetic resonance image.

[0066] Specifically, in this embodiment, during the lesion identification stage, existing algorithms perform high-precision identification of every part of the magnetic resonance image when identifying and labeling lesions, which takes a long time. In this embodiment, two identification models are used to perform "coarse scanning" and "fine scanning" of the magnetic resonance image, respectively. First, the first identification model is used for "coarse scanning". The first identification model uses a lower identification precision to initially identify the areas in the magnetic resonance image that may have lesions (it is not necessary to perform precise boundary labeling of the lesions, it is only necessary to know the approximate range of the lesions). Based on this, the magnetic resonance image is divided into several image blocks (image blocks containing lesions and image blocks not containing lesions) and a first image block set and a second image block set are generated.

[0067] To avoid the "coarse scan" being too inaccurate and missing any lesions, the image blocks in the second image block set are scanned again using a first recognition model with slightly higher accuracy than the first "coarse scan". Based on the scan results, image blocks containing lesions are selected and added to the first image block set.

[0068] Next, a high-precision lesion identification model is used to perform a "precision scan" on each image block containing a lesion in the first image block set, and the lesion boundaries are marked to obtain multiple marked image blocks. After all the markings are completed, all image blocks (each marked image block and the image blocks in the second image set that do not contain lesions) are stitched together (the stitching method is to stitch together according to the positions of the aforementioned segmentation) to obtain the lesion-marked MRI image and display it to the physician and the patient. When displaying the lesion-marked MRI image, each marked image block can also be displayed as a separate part, which makes it easier for the physician to observe and diagnose each lesion.

[0069] Using the above-mentioned method of first "coarse scanning" and then "fine scanning" to annotate lesions in magnetic resonance images can significantly improve the scanning speed of lesions and save time in obtaining lesion-annotated magnetic resonance images.

[0070] In a preferred embodiment of the present invention, a lesion identification method based on magnetic resonance imaging is applied to the aforementioned lesion identification system, such as... Figure 2 As shown, lesion identification methods include:

[0071] Step S1: When the current patient enters the scanning area, the lesion recognition system controls the acquisition coil to scan the area to be scanned of the current patient to obtain the magnetic resonance imaging signal. After the electromagnetic interference of the magnetic resonance imaging signal is eliminated, the image is reconstructed to obtain the magnetic resonance image.

[0072] In step S2, the lesion recognition system inputs a pre-trained first lesion recognition model into the magnetic resonance image, and divides the magnetic resonance image into multiple image blocks according to the recognition results to generate a first image block set and a second image block set. Then, the second image block set is input into the first lesion recognition model and the first and second image block sets are adjusted according to the recognition results. Then, the first image block set is input into the pre-trained second lesion recognition model to identify the lesions and obtain multiple corresponding labeled image blocks. The image blocks in the first image block set and the labeled image blocks are stitched together to obtain the lesion-labeled magnetic resonance image.

[0073] In a preferred embodiment of the present invention, an interference receiving coil is further provided around the scanning area, such as... Figure 3 As shown, step S1 includes:

[0074] Step S11: When the current patient enters the scanning area, the lesion recognition system controls the acquisition coil to scan the area to be scanned of the current patient to obtain magnetic resonance imaging signals.

[0075] In step S12, the lesion identification system controls the interference receiving coil to collect electromagnetic interference signals from interference sources in the space to the scanning area. Then, the electromagnetic interference signals and magnetic resonance imaging signals are input into the pre-trained interference cancellation model to obtain the corresponding cancellation signals as magnetic resonance imaging signals.

[0076] Step S13: The lesion identification system uses an image reconstruction algorithm to reconstruct the image based on the magnetic resonance imaging signal to obtain a magnetic resonance image.

[0077] In a preferred embodiment of the present invention, such as Figure 3 As shown, step S1 further includes:

[0078] Step S14: The lesion identification system uses a mean filtering algorithm, or a median filtering algorithm, or a non-local mean denoising algorithm, or a wavelet denoising algorithm to denoise the magnetic resonance image and then uses it as the magnetic resonance image.

[0079] Step S15: The lesion identification system performs artifact correction processing on the magnetic resonance image and then uses it as the magnetic resonance image.

[0080] Step S16: The lesion identification system performs image enhancement processing on the magnetic resonance image and then uses it as the magnetic resonance image.

[0081] In a preferred embodiment of the present invention, such as Figure 4 As shown, step S2 includes:

[0082] Step S21: The lesion recognition system inputs the magnetic resonance image into the pre-trained first recognition model, and divides the magnetic resonance image into multiple image blocks containing lesions and multiple image blocks not containing lesions according to the recognition results. Then, it selects each image block containing lesions to generate a first image set, and generates a second image set according to the remaining image blocks.

[0083] In step S22, the lesion recognition system inputs each image block in the second image set into the first recognition model, and selects image blocks containing lesions from the second image block set according to the recognition results and adds them to the first image block set.

[0084] In step S23, the lesion identification system inputs the first image block set into the second identification module to mark the boundaries of the lesions contained in each image block in the first image block set to obtain the corresponding marked image blocks. Then, the image blocks and the image blocks in the first image block set are stitched together to obtain the lesion marked magnetic resonance image.

[0085] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A magnetic resonance based lesion identification system, characterized in that, The method comprises the following steps: An image reconstruction module is configured to control a scanning coil to scan a part of a current patient to be scanned to obtain a magnetic resonance imaging signal, and to reconstruct an image based on the magnetic resonance imaging signal after electromagnetic interference elimination to obtain a magnetic resonance image; A lesion identification module is connected to the image reconstruction module and configured to input the magnetic resonance image into a pre-trained first lesion identification model to obtain an identification result, and to segment the magnetic resonance image into a plurality of image blocks based on the identification result to generate a first image block set and a second image block set, the first image block set comprising a plurality of image blocks containing lesions, and the second image block set comprising a plurality of image blocks not containing lesions, then input the second image block set into the first lesion identification model and adjust the first image block set and the second image block set based on the identification result, then input the first image block set into a pre-trained second lesion identification model to identify lesions to obtain a plurality of corresponding identified image blocks, and perform image splicing on each of the image blocks in the first image block set and each of the identified image blocks to obtain a lesion-identified magnetic resonance image.

2. The lesion identification system of claim 1, wherein, An interference receiving coil is further arranged around the scanning area, and the image reconstruction module comprises: A signal acquisition module is configured to control a scanning coil to scan a part of a current patient to be scanned to obtain the magnetic resonance imaging signal; An interference elimination unit is connected to the signal acquisition module and configured to control the interference receiving coil to collect electromagnetic interference signals of a disturbance source in a space around the scanning area, then input the electromagnetic interference signals and the magnetic resonance imaging signal into a pre-trained interference elimination model to obtain corresponding elimination signals as the magnetic resonance imaging signal; An image reconstruction unit is connected to the interference elimination unit and configured to reconstruct an image based on the magnetic resonance imaging signal to obtain the magnetic resonance image.

3. The lesion identification system of claim 2, wherein, The image reconstruction module further comprises: An image denoising unit is connected to the image reconstruction unit and configured to perform image denoising on the magnetic resonance image based on a mean filter algorithm, a median filter algorithm, a non-local mean denoising algorithm, or a wavelet denoising algorithm to obtain the magnetic resonance image; An artifact correction unit is connected to the image denoising unit and configured to perform artifact correction on the magnetic resonance image to obtain the magnetic resonance image; An image enhancement unit is connected to the artifact correction unit and configured to perform image enhancement on the magnetic resonance image to obtain the magnetic resonance image.

4. The lesion identification system of claim 1, wherein, The lesion identification module comprises: A first identification unit is configured to input the magnetic resonance image into the pre-trained first lesion identification model, segment the magnetic resonance image into a plurality of image blocks containing lesions and a plurality of image blocks not containing lesions based on the identification result, then filter out each of the image blocks containing lesions to generate the first image block set, and generate the second image block set based on each of the remaining image blocks. A second identification unit connected to the first identification unit is configured to input each of the image blocks in the second image block set into the first lesion identification model, and to screen the image blocks containing lesions from the second image block set according to the identification result, and add the image blocks containing lesions into the first image block set. A lesion marking unit connected to the second identification unit is configured to input the first image block set into the second lesion identification model to mark the boundaries of the lesions contained in each of the image blocks in the first image block set to obtain corresponding marking image blocks, and then perform image splicing on each of the image blocks and each of the image blocks in the first image block set to obtain the lesion marking magnetic resonance image.

5. A method for lesion identification based on magnetic resonance, characterized in that, The lesion identification method applied to the lesion identification system of any one of claims 1-4 comprises: Step S1, the lesion identification system controls the acquisition coil to scan the to-be-scanned part of the current patient to obtain a magnetic resonance imaging signal when the current patient enters a scanning area, and performs image reconstruction on the magnetic resonance imaging signal after electromagnetic interference elimination to obtain a magnetic resonance image; Step S2, the lesion identification system inputs the magnetic resonance image into a pre-trained first lesion identification model, and according to the identification result, divides the magnetic resonance image into a plurality of image blocks and generates a first image block set and a second image block set, then inputs the second image block set into the first lesion identification model and adjusts the first image block set and the second image block set according to the identification result, then inputs the first image block set into a pre-trained second lesion identification model for lesion marking to obtain a plurality of corresponding marking image blocks, and performs image splicing on each of the image blocks in the first image block set and each of the marking image blocks to obtain a lesion marking magnetic resonance image.

6. The lesion identification method of claim 5, wherein, The scanning area is also provided with an interference receiving coil, and the step S1 comprises: Step S11, the lesion identification system controls the acquisition coil to scan the to-be-scanned part of the current patient to obtain the magnetic resonance imaging signal when the current patient enters the scanning area; Step S12, the lesion identification system controls the interference receiving coil to collect electromagnetic interference signals of the scanning area caused by the interference source in the space, and then inputs the electromagnetic interference signals and the magnetic resonance imaging signal into a pre-trained interference elimination model to obtain corresponding elimination signals as the magnetic resonance imaging signal; Step S13, the lesion identification system uses an image reconstruction algorithm to perform image reconstruction on the magnetic resonance imaging signal to obtain the magnetic resonance image.

7. The lesion identification method of claim 6, wherein, The step S1 further comprises: Step S14, the lesion identification system uses a mean filtering algorithm, or a median filtering algorithm, or a non-local mean denoising algorithm, or a wavelet denoising algorithm to perform image denoising on the magnetic resonance image as the magnetic resonance image; Step S15, the lesion identification system performs artifact correction processing on the magnetic resonance image as the magnetic resonance image; Step S16, the lesion identification system performs image enhancement processing on the magnetic resonance image as the magnetic resonance image.

8. The lesion identification method of claim 5, wherein, The step S2 comprises: Step S21, the lesion recognition system inputs the magnetic resonance image into a pre-trained first lesion recognition model, segments the magnetic resonance image into a plurality of image blocks containing lesions and a plurality of image blocks not containing lesions according to the recognition result, then filters out each image block containing lesions to generate the first image block set, and generates a second image block set according to the remaining image blocks; Step S22, the lesion recognition system inputs each image block in the second image block set into the first lesion recognition model, and filters out the image blocks containing lesions from the second image block set according to the recognition result to add to the first image block set; Step S23, the lesion recognition system inputs the first image block set into the second lesion recognition model to identify the boundaries of the lesions contained in each image block in the first image block set to obtain the corresponding identified image block, then splices each image block and each image block in the first image block set to obtain the lesion identification magnetic resonance image.

Images