Neovascular Tumor Recognition System Based on Image Recognition

Abstract

The present invention discloses a neurovascular tumor recognition system based on image recognition, belonging to the technical field of neurovascular tumor image recognition. It includes an image acquisition and processing module, a feature extraction and evaluation module, and a result visualization module. The feature extraction and evaluation module includes a basic recognition and evaluation unit, a combined hemodynamics evaluation unit, and a malignant risk evaluation unit. The present invention can comprehensively reflect the basic image features of tumors, the growth environment and metabolic requirements of tumors, and integrate multi-dimensional information to accurately evaluate the malignant risk of tumors and reflect their dynamic changes over time, solving the technical problems of single feature evaluation, lack of blood flow information fusion, and failure to consider the dynamic changes in tumor angiogenesis in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of neurovascular tumor image recognition, and particularly to a neurovascular tumor recognition system based on image recognition. Background Art

[0002] In the medical field, the accurate recognition and malignant risk assessment of neurovascular tumors are crucial for formulating treatment plans and judging the prognosis of patients. With the development of medical imaging technology, tumor diagnosis methods based on image recognition have gradually become a research hotspot. These methods analyze medical images to extract characteristic information of tumors to assist doctors in diagnosis and assessment.

[0003] However, existing image recognition systems still have deficiencies in comprehensively considering various characteristics and dynamic changes of tumors, and it is difficult to accurately and comprehensively evaluate the malignant risk of tumors. Secondly, blood flow conditions are closely related to the growth and metabolism of tumors, while most existing systems do not fully consider the hemodynamic information around tumors and cannot accurately evaluate the degree of association between tumors and surrounding blood flow. In addition, existing technologies usually perform static analysis on single images and do not consider the dynamic changes of tumors over time. Specifically, during the treatment process of tumors, multiple detections and identifications are required to observe the development trend of tumors. Among them, the growth, angiogenesis, and metabolism of tumors will change over time, but the evaluation based on the results of a single examination by existing technologies will lead to misjudgment. Moreover, existing technologies will generally calculate the newly generated blood vessels that expand over time within the tumor characteristics. However, benign tumors and malignant tumors have different characteristics in angiogenesis. Malignant tumors usually have stronger angiogenesis ability, and the morphology, quantity, and distribution of newly generated blood vessels are more disordered. Existing technologies do not consider this factor separately, resulting in a reduction in the timeliness and effectiveness of identifying tumor risks. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: there are disadvantages in the prior art such as single - feature evaluation, lack of blood flow information fusion, and failure to consider the dynamic changes in tumor angiogenesis. For this reason, we propose a neurovascular tumor recognition system based on image recognition.

[0005] The technical solution mainly is: a neurovascular tumor recognition system based on image recognition, including an image acquisition and processing module, a feature extraction and evaluation module, and a result visualization module. The feature extraction and evaluation module includes a basic recognition and evaluation unit, a combined hemodynamics evaluation unit, and a malignant risk evaluation unit;

[0006] The specific implementation steps are as follows:

[0007] Step 1: Use the image acquisition and processing module to scan the patient and obtain image data containing various characteristics of the neurovascular tumor, including tomographic images, soft tissue images, metabolic information of the tumor, blood flow velocity and flow around the tumor;

[0008] Step 2: Based on the image data, and using the feature extraction and evaluation module, calculate the tumor basic feature index LT, blood flow correlation feature index X, and tumor malignancy risk index EF in sequence;

[0009] Step 3: For patients undergoing image recognition for the first time, store the data collected and calculated by the image acquisition and processing module and the feature extraction and evaluation module in the image acquisition and processing module, and use the metabolic information of the tumor as the preliminary image recognition result;

[0010] For patients undergoing image recognition for the second time and above, use the result visualization module to draw a linear trend graph of the tumor malignancy risk index EF for image recognition observation.

[0011] Preferably, the devices used by the image acquisition and processing module include a CT scanner, an MRI device, a PET scanner, and an ultrasonic Doppler device. The devices used by the feature extraction and evaluation module include a computer device. The devices used by the result visualization module include a drawing and display device.

[0012] Preferably, the calculation formula of the basic recognition and evaluation unit is as follows:

[0013] ;

[0014] Where:

[0015] LT is the tumor basic feature index;

[0016] M is the tumor area. M measures the number of pixels in the tumor area through image segmentation technology and is converted into the actual area according to the pixel resolution of the image;

[0017] HM is the tumor gray density. HM reflects the average degree of pixel gray values in the tumor area. The specific calculation formula is: ;

[0018] HM i is the gray value of the i-th pixel, and n is the total number of pixels in the tumor area;

[0019] WT is the tumor texture feature value. WT reflects the texture complexity inside the tumor. The specific calculation formula is: ;

[0020] N gis the number of gray levels of the image, and P(i, j) is the element value at the position (i, j) in the gray-level co-occurrence matrix;

[0021] N g-1 represents the number of gray levels N of the image g The corresponding maximum gray-level value index. If the number of gray levels N of the image g is 256, then the gray-level value range is from 0 to 255. At this time, N g-1 is 255;

[0022] B is the gray-level contrast difference value. B is obtained by calculating the difference in the average gray-level values of the tumor region and the surrounding normal tissue region. The specific calculation formula is: ;

[0023] ZM is the surrounding gray-level density;

[0024] NO is the noise level. NO is measured by calculating the standard deviation of the gray-level values of the image background region. The specific calculation formula is: ;

[0025] m is the total number of pixels in the background region, BM i is the gray-level value of the i-th pixel in the background region, and BHM is the average gray-level value of the background region.

[0026] Preferably, the calculation formula of the combined hemodynamics evaluation unit is as follows:

[0027] ;

[0028] Among them:

[0029] X is the blood flow correlation feature index. X reflects the mutual relationship between the tumor and the surrounding blood flow to evaluate the growth and metabolic activity of the tumor;

[0030] V is the average blood flow velocity. V reflects the average blood flow velocity around the tumor;

[0031] LL is the unit blood flow. LL reflects the unit time blood flow around the tumor;

[0032] The result of V + LL is used to comprehensively reflect the hemodynamic state of the blood flow around the tumor;

[0033] JL is the spacing. JL reflects the shortest distance between the blood flow measurement point and the tumor edge, and the blood vessels that provide the main blood supply to the tumor are identified through medical imaging techniques, and blood flow measurement points are selected on these blood vessels.

[0034] Preferably, the calculation formula of the malignant risk evaluation unit is as follows:

[0035] ;

[0036] Wherein:

[0037] EF is the tumor malignancy risk index;

[0038] XM is the angiogenesis index. XM reflects the density of the number of newly formed blood vessels around the tumor through image recognition technology. The specific calculation formula is: XM = GL / GM;

[0039] GL is the number of newly formed blood vessels, and GM is the area of newly formed blood vessels;

[0040] D is the tumor metabolic rate;

[0041] If the result of D is between 2 and 2.5, it is initially identified as a benign tumor;

[0042] If the result of D exceeds 2.5, it is initially identified as a malignant tumor;

[0043] T is the time interval quantity, and T reflects the time interval between two adjacent recognition detections.

[0044] Preferably, the specific recognition and calculation process of the angiogenesis index XM is as follows:

[0045] Step 1: Use a deep learning network to perform pixel-level segmentation on the image data and segment the blood vessel area in the image from the background area;

[0046] Step 2: Use image registration technology to align the image data at different time points, that is, the current and the previous image data, and analyze the morphological and positional characteristics of the blood vessels to determine the number GL of newly formed blood vessels and the area GM of newly formed blood vessels;

[0047] Step 3: Calculate the angiogenesis index XM using XM = GL / GM.

[0048] Preferably, when the angiogenesis index XM is calculated for the first time, it is automatically set to 1;

[0049] When the time interval quantity T is calculated for the first time, it is set in real time by the average time interval between two adjacent recognition detections of patients who have been diagnosed with neurovascular tumors.

[0050] Preferably, the image recognition trend analysis of the line graph drawn by the tumor malignancy risk index EF is as follows:

[0051] If the tumor malignancy risk index EF shows a continuous upward trend on the line graph, it reflects a high risk of malignant development of the tumor;

[0052] If the tumor malignancy risk index EF shows a downward and flattening trend on the line graph, the risk of the tumor tends to be benign.

[0053] Technical effects and advantages of the present invention:

[0054] In the present invention, through the basic recognition and evaluation unit, multiple parameters of the tumor surface M, tumor gray density HM, tumor texture feature value WT, and gray contrast difference value B are comprehensively calculated to obtain the tumor basic feature index LT. This multi-feature fusion method can comprehensively and accurately reflect the image features of the tumor. Based on this, when evaluating the tumor, not only the size of the tumor is considered, but also its internal structure and external boundary features are combined, making the evaluation result more reliable.

[0055] In the present invention, the combined hemodynamics evaluation unit combines the tumor basic feature index LT obtained by image recognition with the blood flow velocity and flow information around the tumor to calculate the blood flow correlation feature index X. This cross-modal information fusion method can more comprehensively reflect the mutual relationship between the tumor and the surrounding blood flow, providing a new perspective for the image recognition of the tumor and improving the evaluation accuracy of the tumor growth environment and metabolic requirements.

[0056] In the present invention, taking the tumor malignancy risk index EF obtained from the first calculation as the original data, by comparing the number GL and area GM of newly generated blood vessels around the tumor each time with the previous time, the blood vessel angiogenesis index around the tumor is updated, and the tumor malignancy risk index EF is recalculated. This dynamic monitoring method can timely detect the changes of the tumor and can also more accurately evaluate the change of the tumor malignancy risk over time. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is the method flow chart of this neurovascular tumor recognition system;

[0058] Figure 2 is the schematic diagram of the recognition and calculation process of the blood vessel angiogenesis index XM in this neurovascular tumor recognition system. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] Now, the present invention will be further described in detail with reference to the accompanying drawings and preferred embodiments.

[0060] Refer to Figure 1 and Figure 2 As shown, the present invention provides a technical solution: a neurovascular tumor recognition system based on image recognition, including an image acquisition and processing module, a feature extraction and evaluation module, and a result visualization module, characterized in that: the feature extraction and evaluation module includes a basic recognition and evaluation unit, a combined hemodynamics evaluation unit, and a malignancy risk evaluation unit;

[0061] The specific implementation steps are as follows:

[0062] Step 1: Use the image acquisition and processing module to scan the patient and obtain image data containing various characteristics of the neurovascular tumor, including tomographic images, soft tissue images, metabolic information of the tumor, blood flow velocity and flow around the tumor;

[0063] Step 2: Based on the image data, and using the feature extraction and evaluation module, calculate the tumor basic feature index LT, blood flow correlation feature index X, and tumor malignancy risk index EF in sequence;

[0064] Step 3: For patients undergoing image recognition for the first time, store the data collected and calculated by the image acquisition and processing module and the feature extraction and evaluation module in the image acquisition and processing module, and use the metabolic information of the tumor as the preliminary image recognition result;

[0065] For patients undergoing image recognition for the second time and above, use the result visualization module to draw a linear trend chart of the tumor malignancy risk index EF for image recognition observation;

[0066] The devices used in the image acquisition and processing module include CT scanners, MRI devices, PET scanners, and ultrasound Doppler devices. The devices used in the feature extraction and evaluation module include computer devices. The devices used in the result visualization module include drawing and display devices.

[0067] In this embodiment, the basic recognition and evaluation unit, combined with the hemodynamic evaluation unit, and the malignancy risk evaluation unit run through the image acquisition and processing module, feature extraction and evaluation module, and result visualization module of the neurovascular tumor recognition system based on image recognition. The basic recognition and evaluation unit calculates the tumor basic feature index LT by comprehensively calculating multiple features in the image feature extraction unit, providing accurate basic image recognition features of the tumor for the system. The hemodynamic evaluation unit combines the image and blood flow information to calculate the blood flow correlation feature index X, realizing cross-modal information fusion, enabling the system to reflect the correlation between the tumor and blood flow. The malignancy risk evaluation unit fuses multi-dimensional information to calculate the tumor malignancy risk index EF, realizing dynamic risk assessment. The combination of the three enables the system to comprehensively, accurately, and dynamically evaluate the malignancy risk of neurovascular tumors with the help of CT, MRI, and ultrasound Doppler devices.

[0068] Refer to Figure 1 As shown, in this implementation scheme: The calculation formula of the basic recognition and evaluation unit is as follows:

[0069] ;

[0070] Where:

[0071] LT is the tumor basic feature index;

[0072] Let \(M\) be the tumor area. \(M\) measures the number of pixels in the tumor region through image segmentation technology and is converted into the actual area according to the pixel resolution of the image;

[0073] Let \(HM\) be the tumor gray density. \(HM\) reflects the average degree of pixel gray values in the tumor region. The specific calculation formula is: ;

[0074] \(HM\) i is the gray value of the \(i\)-th pixel, and \(n\) is the total number of pixels in the tumor region;

[0075] Let \(WT\) be the tumor texture feature value. \(WT\) reflects the texture complexity inside the tumor. The specific calculation formula is:

[0076] ;

[0077] \(N\) g is the number of gray levels of the image, and \(P(i, j)\) is the element value at the position \((i, j)\) in the gray-level co-occurrence matrix;

[0078] \(N\) g-1 represents the index of the maximum gray value corresponding to the number of gray levels \(N\) of the image. If the number of gray levels \(N\) g of the image is 256, then the gray value range is from 0 to 255. At this time, \(N\) g is 255; g-1

[0079] Let \(B\) be the gray contrast difference value. \(B\) is obtained by calculating the difference in the average gray values between the tumor region and the surrounding normal tissue region. The specific calculation formula is:

[0080] ;

[0081] Let \(ZM\) be the surrounding gray density;

[0082] Let \(NO\) be the noise level. \(NO\) is measured by calculating the standard deviation of the gray values in the image background region. The specific calculation formula is:

[0083] ;

[0084] \(m\) is the total number of pixels in the background region, \(BM\) i is the gray value of the \(i\)-th pixel in the background region, and \(BHM\) is the average gray value of the background region.

[0085] In the basic recognition and evaluation unit in this embodiment, The calculation part comprehensively considers the size and internal gray-scale density characteristics of the tumor to reflect the overall quality and internal structural compactness of the tumor. The tumor area M reflects the occupied range of the tumor in the image, and the tumor gray-scale density HM reflects the average level of pixel gray-scale inside the tumor. The multiplication of the two can combine the size and internal structure information of the tumor. Tumors with large area and high gray-scale density have more obvious characteristics. The calculation part then combines the texture characteristics of the tumor and the contrast with the surrounding tissues to describe the external characteristics and boundary clarity of the tumor. The tumor texture characteristic value WT reflects the complexity of the texture inside the tumor, and the gray-scale contrast difference value B reflects the degree of differentiation between the tumor and the surrounding normal tissues. The multiplication of the two can highlight the characteristics of the tumor in terms of texture and boundary, making tumors with complex texture and high contrast with the surrounding tissues easier to identify and distinguish. Thus, The calculation part and The calculation part together constitute the characteristic information of the tumor to more comprehensively reflect the image characteristics of the tumor;

[0086] In the overall calculation part, The calculation part is to normalize the comprehensive tumor characteristic value, that is, the result of the calculation part, to eliminate the influence of image noise on feature evaluation. Among them, the noise level NO will interfere with the accurate judgment of tumor characteristics. Dividing the comprehensive tumor characteristic value by the noise level NO can make the characteristic value comparable in different noise environments and highlight the true characteristics of the tumor. And the square root operation can perform a non-linear transformation on the previously obtained characteristic value to avoid the characteristic value being too large or too small, making the value range of the tumor basic characteristic index LT more in line with the actual situation and facilitating subsequent comparison and analysis.

[0087] Referring to Figure 1 as shown, in this implementation plan: the calculation formula combined with the hemodynamic evaluation unit is as follows:

[0088] ;

[0089] Among them:

[0090] X is the blood flow correlation characteristic index. X evaluates the growth and metabolic activity of the tumor by reflecting the mutual relationship between the tumor and the surrounding blood flow;

[0091] V is the average blood flow velocity. V reflects the average blood flow velocity around the tumor;

[0092] LL is the unit blood flow. LL reflects the unit time blood flow around the tumor;

[0093] The result of V + LL is used to comprehensively reflect the hemodynamic state of the blood flow around the tumor;

[0094] JL is the spacing, which reflects the shortest distance between the blood flow measurement point and the tumor edge. The blood vessels that provide the main blood supply to the tumor are identified through medical imaging techniques, and blood flow measurement points are selected on these blood vessels.

[0095] In the V+LL calculation part of the combined hemodynamics evaluation unit in this embodiment, the velocity and flow rate of the blood flow around the tumor are comprehensively considered to reflect the hemodynamic state around the tumor. The average blood flow velocity V reflects how fast the blood flows in the blood vessel, and the unit blood flow LL of the blood flow reflects the volume of blood passing through a certain cross-section per unit time. Adding the two can more comprehensively describe the blood flow situation around the tumor. A higher velocity and flow rate mean that the tumor has a higher metabolic demand and growth activity, and the calculation part considers the influence of the distance between the blood flow measurement point and the tumor edge on the association between the blood flow and the tumor, and then obtains a ratio that reflects the degree of spatial association between the blood flow and the tumor. The closer the spacing JL is, the greater the influence of the blood flow information on the tumor. The farther the spacing JL is, the smaller the direct association between the blood flow information and the tumor. Dividing the V+LL calculation part by the spacing JL can quantify this degree of spatial association;

[0096] Overall the calculation part comprehensively considers the image features and hemodynamic information of the tumor, and then obtains an index that comprehensively reflects the association between the tumor and the blood flow. Multiplying the calculation part by the tumor basic feature index LT can fuse the image features and blood flow information of the tumor to more accurately evaluate the interaction between the tumor and the surrounding blood flow;

[0097] It should be noted that the closest distance between the tumor and the main blood supply vessel, that is, the spacing JL, can most directly reflect the blood supply situation of the blood vessel to the tumor. The closer the spacing JL is, the shorter the path for the blood to be transported from the blood vessel to the tumor, and the more direct and efficient the supply of nutrients and oxygen from the blood vessel to the tumor, and the greater the impact on the growth and metabolism of the tumor. For the hemodynamic impact, the hemodynamic changes at the closest distance are most closely related to the tumor. Measuring the average blood flow velocity V and the unit blood flow LL of the blood flow at this position can more accurately reflect the blood demand and metabolic state of the tumor. When the tumor grows, its demand for blood increases, and the blood flow in the blood vessel at the closest distance will change accordingly. Thus, as the spacing JL changes, the results reflected by its impact on the blood flow association characteristic index X can also be directly shown.

[0098] Refer to Figure 1 and Figure 2 As shown, in this implementation plan: the calculation formula of the malignant risk assessment unit is as follows:

[0099] ;

[0100] Where:

[0101] EF is the tumor malignancy risk index;

[0102] XM is the angiogenesis index. XM reflects the density of newly formed blood vessels around the tumor through image recognition technology. When the angiogenesis index XM is calculated for the first time, it will be automatically set to 1. The specific calculation formula is: XM = GL / GM;

[0103] GL is the number of newly formed blood vessels, and GM is the area of newly formed blood vessels;

[0104] The specific recognition and calculation process of the angiogenesis index XM is as follows:

[0105] Step 1: Use a deep learning network to perform pixel-level segmentation on the image data and segment the blood vessel area in the image from the background area;

[0106] Step 2: Use image registration technology to align the image data at different time points, that is, the current and the previous image data, and analyze the morphological and positional characteristics of the blood vessels to determine the number GL of newly formed blood vessels and the area GM of newly formed blood vessels;

[0107] Step 3: Calculate the angiogenesis index XM using XM = GL / GM;

[0108] D is the tumor metabolic rate;

[0109] If the result of D is between 2 and 2.5, it is initially identified as a benign tumor;

[0110] If the result of D exceeds 2.5, it is initially identified as a malignant tumor;

[0111] T is the time interval quantity. T reflects the time interval between two adjacent recognition detections. When the time interval quantity T is calculated for the first time, it is set in real time with the average time interval between two adjacent recognition detections of patients diagnosed with neurovascular tumors.

[0112] In the evaluation of the malignancy risk unit of this embodiment The calculation part combines the angiogenesis index around the tumor and the metabolic rate of the tumor to reflect the growth activity and metabolic demand of the tumor. The angiogenesis index XM reflects the number and density of newly formed blood vessels around the tumor. The newly formed blood vessels provide nutrition and oxygen for the tumor, promoting the growth and metastasis of the tumor. The tumor metabolic rate D reflects the metabolic activity degree of tumor cells. The multiplication of the two can combine the angiogenesis and metabolic conditions of the tumor and more comprehensively describe the growth activity of the tumor;

[0113] The calculation part takes into account the impact of the tumor growth time on its malignancy risk, and then obtains a ratio related to the tumor growth time. The longer the time interval T is, the lower the tumor risk is, and the lower the frequency of identification and detection required by the doctor is. Dividing the angiogenesis and metabolism information by the tumor growth time can, to a certain extent, reflect the changes in the tumor growth rate and malignancy risk over time. After taking the square root of this ratio and multiplying it by the blood flow correlation feature index X, the tumor malignancy risk index EF can comprehensively consider various factors such as the tumor image features, blood flow information, angiogenesis, metabolic rate, and growth time to comprehensively evaluate the malignancy risk of the tumor;

[0114] It should be noted that continuous image processing and recognition can capture the early subtle changes in angiogenesis around the tumor. In the initial stage of tumor development, angiogenesis is not obvious, but through continuous monitoring, even the appearance of a small number of new blood vessels can be detected in a timely manner, which helps to make an accurate diagnosis when the tumor is still in the early stage. In addition, the angiogenesis situation is closely related to the nature of the tumor. Malignant tumors usually have a more active angiogenesis ability to meet their rapid growth and metabolic needs. By continuously processing and recognizing the angiogenesis index around the tumor through image processing, it is possible to more accurately judge whether the tumor is benign or malignant.

[0115] Refer to Figure 1 As shown, in this implementation scheme: the image recognition trend analysis of the line graph drawn by the tumor malignancy risk index EF is as follows:

[0116] If the tumor malignancy risk index EF shows a continuous upward trend on the line graph, it reflects a high risk of malignant development of the tumor;

[0117] If the tumor malignancy risk index EF shows a downward and flattening trend on the line graph, the risk of the tumor tends to be benign.

[0118] In this embodiment, by plotting the tumor malignancy risk index EF calculated each time into a line graph, it is possible to intuitively see the changing trend of the tumor malignancy risk over time. Whether the risk gradually increases, decreases, or remains stable can be clearly seen at a glance. Among them, if the line graph shows an upward trend, it indicates that the malignancy risk of the tumor is increasing, suggesting that treatment needs to be strengthened or the treatment plan adjusted. If it shows a downward trend, it means that the current treatment is effective and the tumor condition has improved. In addition, the line graph can capture the subtle changes in the tumor malignancy risk index EF during continuous observation. Even if the change amplitude of each index is small, it can be clearly shown in the continuous line graph. These subtle changes will be the early signals of tumor development and treatment effect.

[0119] It should be noted that any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall also be within the protection scope of the present invention.

Claims

1. A neurovascular tumor recognition system based on image recognition, comprising an image acquisition and processing module, a feature extraction and evaluation module, and a result visualization module, characterized in that: The feature extraction and evaluation module includes a basic recognition and evaluation unit, a combined hemodynamics evaluation unit, and a malignant risk evaluation unit; The specific implementation steps are as follows: Step 1: Using the image acquisition and processing module, scan the patient and obtain image data containing various features of the neurovascular tumor, including tomographic images, soft tissue images, tumor metabolic information, blood flow velocity and flow around the tumor; Step 2: Based on the image data and using the feature extraction and evaluation module, calculate the tumor basic feature index LT, the blood flow correlation feature index X, and the tumor malignant risk index EF in sequence; Step 3: For patients undergoing image recognition for the first time, store the data collected and calculated by the image acquisition and processing module and the feature extraction and evaluation module in the image acquisition and processing module, and use the tumor metabolic information as the preliminary image recognition result; For patients undergoing image recognition for the second time and above, use the result visualization module to draw a linear trend graph of the tumor malignant risk index EF for image recognition observation; The calculation formula of the basic recognition and evaluation unit is as follows: ； Where: LT is the tumor basic feature index; M is the tumor area. M measures the number of pixels in the tumor area through image segmentation technology and converts it into the actual area according to the pixel resolution of the image; HM is the tumor gray density, which reflects the average degree of pixel gray values in the tumor area; WT is the tumor texture feature value, which reflects the texture complexity inside the tumor; B is the gray contrast difference value, which is obtained by calculating the difference in the average gray values of the tumor area and the surrounding normal tissue area; NO is the noise level, which is measured by calculating the standard deviation of the gray values in the image background area; The calculation formula of the combined hemodynamics evaluation unit is as follows: ； Where: X is the blood flow correlation feature index, which reflects the mutual relationship between the tumor and the surrounding blood flow to evaluate the growth and metabolic activity of the tumor; V is the average blood flow velocity, which reflects the average blood flow velocity around the tumor; LL is the unit blood flow, which reflects the unit time blood flow around the tumor; The result of V + LL is used to comprehensively reflect the hemodynamic state of the blood flow around the tumor; JL is the spacing, which reflects the shortest distance between the blood flow measurement point and the tumor edge, and identifies the blood vessels that provide the main blood supply to the tumor through medical imaging technology and selects blood flow measurement points on these blood vessels; The calculation formula of the malignant risk evaluation unit is as follows: ； Where: EF is the tumor malignant risk index; XM is the angiogenesis index, which reflects the density of the number of newly formed blood vessels around the tumor through image recognition technology. The specific calculation formula is: XM = GL / GM; GL is the number of newly formed blood vessels, and GM is the area of newly formed blood vessels; D is the tumor metabolic rate; If the result of D is between 2 and 2.5, it is initially identified as a benign tumor; If the result of D exceeds 2.5, it is initially identified as a malignant tumor; T is the time interval amount, which reflects the time interval between two adjacent recognition detections.

2. The neurovascular tumor recognition system based on image recognition according to claim 1, wherein: The devices used in the image acquisition and processing module include CT scanners, MRI devices, PET scanners, and ultrasonic Doppler devices. The devices used in the feature extraction and evaluation module include computer devices. The devices used in the result visualization module include drawing and display devices.

3. The neurovascular tumor recognition system based on image recognition according to claim 2, characterized in that: The specific identification and calculation process of the angiogenesis index XM is as follows: Step 1: Use a deep learning network to perform pixel-level segmentation on the image data and segment the blood vessel area in the image from the background area; Step 2: Use image registration technology to align the image data at different time points, that is, the current and the previous image data, and analyze the morphological and positional characteristics of the blood vessels to determine the number GL of new blood vessels and the area GM of new blood vessels of the new blood vessels; Step 3: Use XM = GL / GM to calculate the angiogenesis index XM.

4. The image recognition-based neurovascular tumor recognition system according to claim 3, characterized in that: When the angiogenesis index XM is calculated for the first time, it will be automatically set to 1; When the time interval T is calculated for the first time, the average time interval between two adjacent identifications and detections of patients diagnosed with neurovascular tumors will be updated and set in real time.

5. The neurovascular tumor recognition system based on image recognition according to claim 4, wherein: The image recognition trend analysis of the line graph drawn by the tumor malignancy risk index EF is as follows: If the tumor malignancy risk index EF shows a continuous upward trend on the line graph, it reflects a high risk of malignant development of the tumor; If the tumor malignancy risk index EF shows a downward and flattening trend on the line graph, the risk of the tumor tends to be benign.

Images