A low-latency brain electrophysiological monitoring method and system during DSA surgery

Abstract

The present invention relates to the technical field of electrophysiological monitoring of the brain, and specifically to a low-latency electrophysiological monitoring method and system for the brain in DSA surgery, comprising acquiring continuous multi-frame scan images after contrast agent injection, establishing a mapping relationship between time and image information entropy, and generating an initial population of predicted exposure time series, including the existence state, predicted time value, and predicted information entropy of each exposure time point. The population is optimized through iterative evolution, the mean square error between the predicted information entropy and the actual information entropy is evaluated, and the winning individual is selected for cross-mutation. After each scan, the mapping relationship is dynamically updated, the predicted information entropy is adjusted, and exposure points with too small intervals are eliminated, thereby obtaining the optimal exposure moment sequence. The present invention significantly reduces the number of X-ray exposures during the DSA examination process, thereby greatly reducing the radiation dose received by the patient; by predicting the moment when the contrast agent information entropy is highest and performing X-ray exposure, a clearer vascular image is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of brain electrophysiological monitoring, and in particular to a low-latency brain electrophysiological monitoring method and system during DSA surgery. Background Art

[0002] Digital Subtraction Angiography (DSA) is a medical imaging technique that uses contrast agent injection to visualize blood vessels and then combines computer processing to eliminate interference from bone and soft tissue, resulting in high-contrast dynamic images of blood vessels. Its core principle is to highlight vascular structures through "subtraction" (i.e., comparing images before and after contrast agent injection). For thrombosis diagnosis in brain surgery, DSA currently uses carbon fiber electrodes (non-visible electrodes) to eliminate interference with imaging caused by traditional metal electrodes. During the imaging process, the patient's EEG / SEP signals and other physiological electrical signals are also monitored in real time, and the relevant data is transmitted remotely for processing. During this procedure, contrast agent is injected via an arterial puncture (such as the femoral or radial artery) and a catheter is inserted. Subsequently, multiple dynamic scans are performed during the procedure using X-rays. However, the contrast agent injection rate and exposure timing require manual adjustment by the surgeon, which can lead to inconsistent image quality due to operator variability. Consequently, more X-ray exposures are required to ensure optimal images. Summary of the Invention

[0003] (1) Technical problems to be solved

[0004] The purpose of the present invention is to provide a low-latency brain electrophysiological monitoring method and system during DSA surgery to improve the quality of the DSA imaging process and reduce the number of X-ray exposures as much as possible.

[0005] (2) Technical solution

[0006] To achieve the above objectives, the present invention provides a low-latency electrophysiological brain monitoring method during DSA surgery, the method comprising:

[0007] Acquire multiple consecutive scanned images after contrast agent injection during DSA surgery using a non-developing electrode, and establish an initial diffusion image sequence including the scan time and corresponding image data; obtain image information entropy through the grayscale distribution of each frame image, and establish a mapping relationship between the scan time and image information entropy in the initial diffusion image sequence;

[0008] Generate an initial population for predicting exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes an existence state, a predicted time value, and a predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy.

[0009] The initial population is iteratively evolved. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with state 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold.

[0010] After each scan is completed, the mapping relationship between the scanning time and the image information entropy is updated; the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence is dynamically adjusted according to the updated mapping relationship; the minimum time interval threshold between adjacent exposure time points is set, and the exposure time points with a time interval less than the threshold are eliminated; the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached are used as subsequent exposure moments.

[0011] Furthermore, the method of acquiring a plurality of consecutive scanned images after contrast agent injection, establishing an initial diffusion image sequence including scanning time and corresponding image data; obtaining image information entropy by grayscale distribution of each frame of image, and establishing a mapping relationship between scanning time and image information entropy in the initial diffusion image sequence includes:

[0012] Get the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area ;No. Contrast agent diffusion information entropy at each scanning moment The calculation method is:

[0013] ;

[0014] in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent;

[0015] No. Vascular imaging quality information entropy at each scanning moment The calculation method is:

[0016] ;

[0017] in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

[0018] Furthermore, the acquisition of the first diffusion image sequence The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area The methods include:

[0019] Get the Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ;

[0020] The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within :

[0021] ;

[0022] in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representative sub-region The number of pixels included;

[0023] Calculate the Contrast agent diffusion intensity at each scanning moment :

[0024] ;

[0025] in Representative The average gray value of all sub-areas at a scanning moment:

[0026] ;

[0027] Calculate the Contrast agent diffusion area at each scanning moment :

[0028] .

[0029] Furthermore, the initial population is iteratively evolved, and the iterative evolution step includes calculating the mean square difference between the predicted information entropy of the exposure time point with state 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation index. The method of selecting a preset number of winning individuals through a roulette wheel selection method based on the fitness index includes:

[0030] The first Individual Represented as a set of forecast exposure time series schemes:

[0031] ;

[0032] in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable;

[0033] No. Predicting exposure time series schemes Fitness evaluation index The calculation method is:

[0034] ;

[0035] in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1;

[0036] According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for:

[0037] ;

[0038] in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

[0039] Furthermore, the method further comprises:

[0040] Get the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ;

[0041] The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

[0042] Based on the same inventive concept, on the other hand, the present invention further provides a low-latency brain electrophysiological monitoring system during DSA surgery, the system comprising:

[0043] An exposure image acquisition module is used to acquire continuous multi-frame scan images after contrast agent injection during DSA surgery using a non-developing electrode, and to establish an initial diffusion image sequence including the scan time and the corresponding image data; the image information entropy is obtained by the grayscale distribution of each frame of the image, and a mapping relationship between the scan time and the image information entropy in the initial diffusion image sequence is established;

[0044] The parameter iteration initialization module is used to generate the initial population of the predicted exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes the existence state, the predicted time value, and the predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy.

[0045] A parameter iteration module is used to iteratively evolve the initial population. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with a state of 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold.

[0046] The dynamic exposure time adjustment module is used to update the mapping relationship between the scanning time and the image information entropy after each scan is completed; dynamically adjust the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence according to the updated mapping relationship; set the minimum time interval threshold between adjacent exposure time points, and eliminate exposure time points with a time interval less than the threshold; and use the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached as the subsequent exposure time.

[0047] Furthermore, the system further comprises:

[0048] X-ray image information entropy calculation module is used to obtain the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area ;No. Contrast agent diffusion information entropy at each scanning moment The calculation method is:

[0049] ;

[0050] in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent;

[0051] No. Vascular imaging quality information entropy at each scanning moment The calculation method is:

[0052] ;

[0053] in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

[0054] Furthermore, the system further comprises:

[0055] Contrast agent diffusion modeling module, used to obtain the first Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ;

[0056] The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within :

[0057] ;

[0058] in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representative sub-region The number of pixels included;

[0059] Calculate the Contrast agent diffusion intensity at each scanning moment :

[0060] ;

[0061] in Representative The average gray value of all sub-areas at a scanning moment:

[0062] ;

[0063] Calculate the Contrast agent diffusion area at each scanning moment :

[0064] .

[0065] Furthermore, the system further comprises:

[0066] Genetic iteration module is used to convert the first Individual Represented as a set of forecast exposure time series schemes:

[0067] ;

[0068] in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable;

[0069] No. Predicting exposure time series schemes Fitness evaluation index The calculation method is:

[0070] ;

[0071] in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1;

[0072] According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for:

[0073] ;

[0074] in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

[0075] Furthermore, the system further comprises:

[0076] Exposure moment fine-tuning module, used to obtain the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ;

[0077] The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

[0078] (3) Beneficial effects

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. Significantly reduce the number of X-ray exposures during DSA examinations, significantly reducing the radiation dose received by patients and improving the safety of the examination.

[0081] 2. By predicting and capturing the moment when the contrast agent information entropy is highest and performing X-ray exposure, clearer vascular images can be obtained, improving accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 This is a flowchart of a low-latency electrophysiological brain monitoring method during DSA surgery according to Example 1 of the present invention;

[0083] Figure 2 This is a module block diagram of a low-latency brain electrophysiological monitoring system during DSA surgery according to Example 2 of the present invention. DETAILED DESCRIPTION

[0084] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0085] Before giving examples, it's necessary to explain the application scenarios of the present invention. Generally, the radiation dose of a single X-ray examination is 0.1 mSv. A single DSA angiography involves capturing a series of contrast agent diffusion maps (synchronously observed in conjunction with physiological EEG signals). Each angiography image is synthesized by subtracting two images (one before contrast agent injection, and one at a certain moment after). Therefore, the radiation dose of a single DSA can reach 20 to 200 mSv (the angiography process requires frequent imaging, and the actual contrast dose depends on the surgeon's skill level). Therefore, rationally reducing the number of X-ray exposures during the DSA process can help reduce the radiation dose of a single DSA.

[0086] Example 1: Figure 1 As shown, this embodiment provides a low-latency brain electrophysiological monitoring method during DSA surgery, the method comprising:

[0087] Acquire multiple consecutive scanned images after contrast agent injection during DSA surgery using a non-developing electrode, and establish an initial diffusion image sequence including the scan time and corresponding image data; obtain image information entropy through the grayscale distribution of each frame image, and establish a mapping relationship between the scan time and image information entropy in the initial diffusion image sequence;

[0088] Generate an initial population for predicting exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes an existence state, a predicted time value, and a predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy.

[0089] The initial population is iteratively evolved. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with state 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold.

[0090] After each scan is completed, the mapping relationship between the scanning time and the image information entropy is updated; the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence is dynamically adjusted according to the updated mapping relationship; the minimum time interval threshold between adjacent exposure time points is set, and the exposure time points with a time interval less than the threshold are eliminated; the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached are used as subsequent exposure moments.

[0091] For example, during a hospital patient's routine DSA (Digital Surgery) procedure using X-rays for imaging, this method was used to optimize exposure timing, ensuring strict adherence to relevant regulations throughout the entire process. The patient was a 68-year-old male with cerebral vascular stenosis. Each DSA lasted 45 to 60 minutes, including puncture, catheterization of the arterial cannula, initial X-ray image acquisition, contrast agent injection, and timed X-ray acquisition of contrast images. The patient was 170 cm tall and weighed 85 kg, making him somewhat overweight, and the contrast agent was expected to diffuse slowly.

[0092] First, iohexol contrast agent was injected via femoral artery puncture at a concentration of 370 mg iodine / mL at a rate of 4 mL / s for a total volume of 15 mL. An initial diffusion image sequence was acquired using a biplane angiography device, capturing 20 frames at key time points, ranging from 1 to 40 seconds after contrast injection, with a sampling interval of 2 seconds. This 2-second sampling interval was chosen based on the patient's BMI (29.4) and age to ensure full contrast agent diffusion while minimizing excessive radiation exposure. This empirical data serves only as a variable for the initial exposure time. For each of the 20 scanned images, the grayscale distribution was calculated, yielding the following image information entropy values ​​at each time: 0.87 at 2 seconds, 3.25 at 6 seconds, 4.92 at 10 seconds, 5.83 at 14 seconds, 5.76 at 18 seconds, 5.34 at 22 seconds, 4.85 at 26 seconds, 4.26 at 30 seconds, 3.65 at 34 seconds, and 3.12 at 38 seconds. Cubic spline interpolation was used to establish the mapping between time and information entropy. This method was chosen because the contrast agent diffusion curve for this patient exhibited significant nonlinear characteristics, with complex variations particularly in the peak region.

[0093] Then, an initial population of 40 individuals was generated based on the mapping relationship. Each individual represented a predicted exposure time series scenario, and each scenario contained 10 potential exposure time points. Taking the first individual as an example, the parameter states of the 10 exposure time points were as follows: the first time point had a presence state of 1, a predicted time of 7.5 seconds, and a predicted information entropy of 3.86; the second time point had a presence state of 1, a predicted time of 14.3 seconds, and a predicted information entropy of 5.87; the third time point had a presence state of 0, a predicted time of 18.9 seconds, and a predicted information entropy of 5.63; the fourth time point had a presence state of 1, a predicted time of 24.5 seconds, and a predicted information entropy of 4.73; the fifth time point had a presence state of 1, a predicted time of 32.8 seconds, and a predicted information entropy of 3.42; the sixth to tenth time points were 38.5 seconds, 44.2 seconds, 52.6 seconds, 61.4 seconds, and 70.3 seconds, respectively. The presence states of each point were randomly set, and the predicted information entropy was calculated using the mapping function. The predicted time value for each exposure time point was generated by uniform random sampling based on the time range of the initial diffusion image sequence (1–40 s) and appropriately extended to the range of 70 s, taking into account the continuous diffusion process of the contrast agent in actual situations.

[0094] The initial population was iteratively evolved, and the fitness evaluation metric for each prediction scheme was first calculated. Taking the first individual as an example, actual images were collected at five exposure time points (7.5 seconds, 14.3 seconds, 24.5 seconds, 32.8 seconds, and 52.6 seconds) where the presence state was 1. The measured actual information entropies were 3.98, 5.82, 4.65, 3.51, and 2.28, respectively, with a mean square error (MSE) of 0.0063 compared to the predicted information entropy. The MSE was calculated using a standard formula: the sum of the squares of the differences between the predicted and actual information entropies at each time point divided by the number of time points. Based on the calculated fitness metrics, 20 winning individuals were selected using a roulette wheel selection method. The fitness of the first individual was 0.0063, and the maximum overall fitness was 0.0217. A crossover operation was performed on these 20 winning individuals using a two-point crossover strategy, randomly selecting two crossover points and exchanging the parameters between the individuals located between these two points. The offspring individuals are then mutated with a mutation probability set to 15%. The mutation operation involves randomly changing the existence state or adjusting the prediction time within a range of ±3 seconds. The mutation probability and range parameters are determined based on previous clinical validation, and can provide sufficient search diversity while maintaining population stability. The initial population is replaced with the generated offspring individuals, and the above iterative evolution process is repeated. When the fitness improvement values ​​for three consecutive generations are all lower than the set threshold of 0.0003, the iteration terminates. In this case, the termination condition was reached after 22 iterations, and the mean square error of the final winning solution dropped to 0.0031, which was significantly lower than the initial solution.

[0095] After each actual scan is completed, the mapping relationship between the scan time and the image information entropy is updated in real time. For example, after completing a 7.5-second scan, the actual measured information entropy of 3.98 at that point replaces the original predicted value of 3.86, the cubic spline interpolation parameters are recalculated, and the entire mapping function is updated. Based on the updated mapping relationship, the predicted information entropy of the exposure time points to be collected in the scheme is dynamically adjusted. For example, the predicted information entropy of 14.3 seconds is adjusted from 5.87 to 5.91, and the predicted information entropy of 24.5 seconds is adjusted from 4.73 to 4.77. At the same time, the minimum time interval threshold between adjacent exposure time points is set to 4.5 seconds. Adjacent points with a time interval less than the threshold are checked, and points with higher information entropy are retained. In this case, the interval between 18.9 and 14.3 seconds was found to be 4.6 seconds, close to the threshold but not triggering elimination. However, the interval between 44.2 and 38.5 seconds was found to be 5.7 seconds, while the interval between 38.5 and 32.8 seconds was only 3.8 seconds. Therefore, the 38.5-second point was automatically eliminated, retaining the 32.8-second point with higher information entropy. Ultimately, the time points where the state was 1 and the time value had not yet been reached among the winning individuals after iterative evolution were used as subsequent exposure moments, including key time points such as 14.3, 24.5, 32.8, and 44.2 seconds.

[0096] During this DSA examination, a total of 43 X-rays were performed using the aforementioned optimization method: an initial 20 frames and a subsequent 23 frames optimized. The initial 20 frames were used to establish the time-information entropy mapping relationship; the optimized 23 frames included high-quality imaging at nine key time points and 14 detailed observations of the left internal carotid artery stenosis. The 43 imaging times, rather than the 80-120 required with traditional methods (compared to the number of DSAs performed on patients with similar medical records), were achieved because this method continuously optimizes subsequent imaging times based on real-time image information, ensuring that each imaging session occurs at a time of high information entropy, thus avoiding the redundant imaging required with traditional methods.

[0097] Furthermore, the method of acquiring a plurality of consecutive scanned images after contrast agent injection, establishing an initial diffusion image sequence including scanning time and corresponding image data; obtaining image information entropy by grayscale distribution of each frame of image, and establishing a mapping relationship between scanning time and image information entropy in the initial diffusion image sequence includes:

[0098] Get the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area ;No. Contrast agent diffusion information entropy at each scanning moment The calculation method is:

[0099] ;

[0100] in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent;

[0101] No. Vascular imaging quality information entropy at each scanning moment The calculation method is:

[0102] ;

[0103] in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

[0104] For example, in a DSA examination of a 68-year-old male patient with moderate obesity, the first step is to acquire an image at the third scan time (6 seconds) in the initial diffusion image sequence, followed by a mask image before contrast agent injection. The mask image is a baseline image acquired before contrast agent injection and is used for subsequent subtraction processing. The mask image is subtracted from the 6-second image to produce a difference angiographic image. The subtraction process eliminates the effects of bone and soft tissue, highlighting the contrast agent in vascular structures.

[0105] The contrast agent diffusion intensity and diffusion area were extracted from the brain vascular region using angiographic differential images. Specifically, the average grayscale value of the contrast agent within the vascular region at the 6-second scan time was calculated to be 142.8, representing the concentration of the contrast agent in the blood vessels. The maximum average grayscale value observed during the contrast agent diffusion process was 217.5, occurring at the 14-second scan time. The contrast agent diffusion area within the vascular region at the 6-second scan time was also calculated to be 28,764 pixels, representing the size of the area covered by the contrast agent. The maximum diffusion area observed during the contrast agent diffusion process was 37,892 pixels, occurring at the 18-second scan time.

[0106] Using the measured data above, we calculated the contrast agent diffusion information entropy at the 6-second scan time. Substituting the diffusion intensity ratio of 0.657 and the diffusion area ratio of 0.759 into the relevant formula, we obtained a contrast agent diffusion information entropy of 3.25 at that time. This process primarily considers two key parameters, contrast agent concentration and coverage area, and can comprehensively reflect the contrast agent diffusion state.

[0107] Next, we calculated the vascular imaging quality information entropy at the 6-second scan time: We analyzed the angiographic difference image, identified the 12 major vascular branch regions, and calculated the grayscale gradient value for each region. For example, the grayscale gradient value of the first vascular branch region was 18.4, the second was 21.5, and so on. The sum of the grayscale gradient values ​​of all vascular branch regions was 273.6. Using the relevant formula, we calculated the vascular imaging quality information entropy value at this moment to be 0.876.

[0108] The total image information entropy at the 6-second scan time was calculated as 2.847, using the product of the contrast agent diffusion information entropy (3.25) and the vascular imaging quality information entropy (0.876). The same method was used to calculate the total image information entropy for all 20 scan times, resulting in values ​​of 4.92 at 10 seconds, 5.83 at 14 seconds, and 5.76 at 18 seconds. A mapping function between scan time and image information entropy was established using cubic spline interpolation, which can predict the image information entropy value at any time. Cubic spline interpolation was chosen because the contrast agent diffusion curve for this patient exhibited significant nonlinear characteristics, with complex variations particularly in the peak region between 10 and 20 seconds.

[0109] Taking a 14-second scan as an example, the actual measured total information entropy of the image is 5.83, while the value predicted using cubic spline interpolation is 5.79, with a relative error of only 0.7%, verifying the accuracy of the interpolation method.

[0110] Furthermore, the acquisition of the first diffusion image sequence The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area The methods include:

[0111] Get the Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ;

[0112] The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within :

[0113] ;

[0114] in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representative sub-region The number of pixels included;

[0115] Calculate the Contrast agent diffusion intensity at each scanning moment :

[0116] ;

[0117] in Representative The average gray value of all sub-areas at a scanning moment:

[0118] ;

[0119] Calculate the Contrast agent diffusion area at each scanning moment :

[0120] .

[0121] For example, the DSA examination of a 68-year-old obese male patient is continued. After obtaining the angiography difference image at the third scanning moment (6 seconds), the image needs to be further processed to extract the contrast agent diffusion intensity and diffusion area. First, the angiography difference image at 6 seconds is binarized to obtain a binary mask of the brain vascular area. The binarization process uses an adaptive threshold method, and the threshold is set to 128% of the average grayscale value of the image, that is, the pixels with grayscale values ​​higher than the threshold are marked as vascular areas (value 1), and the remaining pixels are marked as background (value 0). The binarization results show that there are some noise points and discontinuous areas in the initial binary mask, which will affect the accuracy of subsequent analysis.

[0122] Morphological operations are performed on the binary mask to optimize its boundaries, including dilation and erosion, to produce an optimized mask of the brain vascular region. Specifically, two dilation operations are performed using a 3×3 structuring element to fill small voids within the vascular region; then, one erosion operation is performed using the same structuring element to remove edge noise and burrs. This morphological processing takes into account the anatomical characteristics of the brain vessels, effectively preserving vascular contours while reducing noise interference.

[0123] The optimized brain vascular region mask was divided into 15 subregions, with the boundaries of each subregion determined using a region growing algorithm. The region growing algorithm selects seed points by first selecting several high-brightness points on the main vascular trunk as initial seed points. The region is then gradually expanded based on pixel grayscale values ​​and spatial continuity. For example, the seed point for the middle cerebral artery region was set at a pixel with a grayscale value of 210, and the growing threshold was set to a grayscale difference of no more than 15 between adjacent pixels. This region segmentation method can distinguish different vascular branches based on the actual anatomical structure of the vessels.

[0124] The average grayscale value within each subregion was calculated. For example, the average grayscale value within the first subregion (corresponding to the proximal segment of the anterior cerebral artery) was 169.2, encompassing 2827 pixels; the average grayscale value within the second subregion (corresponding to the M1 segment of the middle cerebral artery) was 182.5, encompassing 3164 pixels; and so on. This regional calculation can reflect the distribution of contrast agent in different vascular branches and provide a more detailed characterization of diffusion.

[0125] The contrast agent diffusion intensity was calculated for the 6-second scan. First, the average grayscale value of all subregions was calculated, which was 142.8. The weighted standard deviation of each subregion's grayscale value from the average was then calculated, and the product of this and the average grayscale value was used as the diffusion intensity, resulting in a value of 38947. This metric comprehensively considers the overall concentration and distribution uniformity of the contrast agent. A high diffusion intensity indicates that the contrast agent has fully penetrated the vascular system and provides good contrast.

[0126] The contrast agent diffusion area at the 6-second scan time was calculated, and the pixel counts of all 15 subregions were summed, totaling 28,764 pixels. The contrast agent diffusion intensity and area at the 6-second scan time were accurately extracted. The same processing method was applied to the other 19 scan times to obtain a complete data series of contrast agent diffusion intensity and area. In this patient, this method successfully identified the area of ​​stenosis in the left internal carotid artery, as evidenced by the significantly lower grayscale value of subregion 5 compared to the corresponding area on the contralateral side (subregion 12), with a difference of 23%.

[0127] Furthermore, the initial population is iteratively evolved, and the iterative evolution step includes calculating the mean square difference between the predicted information entropy of the exposure time point with state 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation index. The method of selecting a preset number of winning individuals through a roulette wheel selection method based on the fitness index includes:

[0128] The first Individual Represented as a set of forecast exposure time series schemes:

[0129] ;

[0130] in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable;

[0131] No. Predicting exposure time series schemes Fitness evaluation index The calculation method is:

[0132] ;

[0133] in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1;

[0134] According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for:

[0135] ;

[0136] in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

[0137] For example, the DSA examination of a 68-year-old obese male patient was continued. After calculating the mean square error between the predicted information entropy of the exposure time point with the presence state of 1 in each predicted exposure time series scheme and the information entropy of the actual acquired image, the 18th individual in the initial population was represented as a set of predicted exposure time series schemes. The scheme contains 10 exposure time points, each of which is represented by three parameters: presence state, predicted time value, and predicted information entropy. The specific parameters are as follows: the presence state at time point 1 is 1, the predicted time is 8.2 seconds, and the predicted information entropy is 4.21; the presence state at time point 2 is 1, the predicted time is 15.7 seconds, and the predicted information entropy is 5.88; the presence state at time point 3 is 0, the predicted time is 20.3 seconds, and the predicted information entropy is 5.41; the presence state at time point 4 is 1, the predicted time is 26.8 seconds, and the predicted information entropy is 4.57; the presence state at time point 5 is 1, the predicted time is 34.5 seconds, and the predicted information entropy is 3. 36; the existence state at time point 6 is 0, the predicted time is 40.2 seconds, and the predicted information entropy is 2.87; the existence state at time point 7 is 1, the predicted time is 46.7 seconds, and the predicted information entropy is 2.45; the existence state at time point 8 is 0, the predicted time is 53.4 seconds, and the predicted information entropy is 2.12; the existence state at time point 9 is 1, the predicted time is 60.1 seconds, and the predicted information entropy is 1.84; the existence state at time point 10 is 0, the predicted time is 68.5 seconds, and the predicted information entropy is 1.53.

[0138] For the 18th predicted exposure time series, its fitness evaluation metric was calculated. First, five exposure time points with a presence state of 1 were selected: 8.2 seconds, 15.7 seconds, 26.8 seconds, 46.7 seconds, and 60.1 seconds. Actual images were then collected at these five time points, and the measured information entropies were 4.35, 5.92, 4.42, 2.38, and 1.78, respectively. The squared differences between the predicted and actual information entropies were: (4.21 - 4.35)² = 0.0196, (5.88 - 5.92)² = 0.0016, (4.57 - 4.42)² = 0.0225, (2.45 - 2.38)² = 0.0049, and (1.84 - 1.78)² = 0.0036. The squared sum of these five differences and the resulting mean square error (MSE) was 0.0104.

[0139] The fitness evaluation indicators for the other 39 prediction schemes were calculated using the same method. The 7th scheme had the smallest mean square error (MSE) of 0.0064, while the 32nd scheme had the largest MSE of 0.0352. All 40 schemes were ranked according to the calculated fitness evaluation indicators. The fitness rankings, from highest to lowest (MSE to highest), were: 7th scheme (0.0064), 23rd scheme (0.0078), 18th scheme (0.0104), 11th scheme (0.0125), and so on.

[0140] A preset number of winning individuals are selected using a roulette wheel selection method. The probability of each solution being selected is first calculated. Taking the 18th solution as an example, its probability is calculated as (0.0352 - 0.0104) / (0.0352 × 40 - total mean square error sum) = 0.0248 / (0.0352 × 40 - 0.6245) = 0.0248 / 0.7835 = 0.0317. This means that during the roulette wheel selection process, the 18th solution has a 3.17% probability of being selected. Because the roulette wheel selection method allocates selection probabilities based on fitness ratios rather than simply selecting the top few, even solutions with lower fitness have a certain probability of being retained. This helps maintain population diversity and prevents premature convergence to local optima.

[0141] In this example, the roulette wheel selection process is as follows: a random number between 0 and 1 is generated, such as 0.2735. Then, starting with the first option, the selection probabilities are accumulated. The first time the accumulated value exceeds the random number, the corresponding option is selected. This process is repeated 20 times, resulting in 20 winning individuals. Based on this random selection mechanism, the 7th option is selected three times, the 23rd option is selected twice, the 18th option is selected once, and so on, for a total of 20 winning individuals, with some options being selected repeatedly.

[0142] Furthermore, the method further comprises:

[0143] Get the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ;

[0144] The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

[0145] For example, the DSA examination of a 68-year-old obese male patient is continued. After the roulette wheel selection method is completed to select the winning individuals, the predicted exposure time points of these winning individuals are optimized for time offset to ensure that the intervals between adjacent exposure time points meet the minimum time interval requirements while maintaining a high prediction accuracy.

[0146] In the predicted exposure time sequence scheme of the 23rd winning individual, there were 6 exposure time points with state 1, namely 7.5 seconds, 14.8 seconds, 21.2 seconds, 29.7 seconds, 41.5 seconds and 57.3 seconds. The intervals between adjacent exposure time points were 7.3 seconds, 6.4 seconds, 8.5 seconds, 11.8 seconds and 15.8 seconds, respectively. The minimum time interval threshold between adjacent exposure time points was set to 6 seconds, which is based on the average flow rate of contrast agent in cerebral blood vessels (approximately 12-15 cm / s). Inspection found that all intervals were greater than the threshold, but the two interval values ​​of 7.3 seconds and 6.4 seconds were close to the threshold, indicating that there was room for optimization.

[0147] The time offset threshold of 6 seconds was divided into 12 equal parts, resulting in a time offset step of 0.5 seconds. The predicted time value for each presence state of 1 was adjusted sequentially. First, the second exposure time point of 14.8 seconds was adjusted to 14.8 seconds + 0.5 seconds = 15.3 seconds. This increased the interval from the first exposure time point of 7.5 seconds to 7.8 seconds. The adjusted predicted information entropy calculated using the mapping function was 5.89, slightly higher than the original value of 5.87.

[0148] Next, we tried adjusting the third exposure time (21.2 seconds) to 21.2 seconds + 0.5 seconds = 21.7 seconds, increasing the interval with the second exposure time (15.3 seconds) to 6.4 seconds. The predicted information entropy after adjustment was 5.32, a small change from the original value of 5.35. We continued to try adjusting to 21.2 seconds + 1.0 seconds = 22.2 seconds, increasing the interval to 6.9 seconds. The predicted information entropy was 5.26, still within the acceptable range. We then tried adjusting to 21.2 seconds + 1.5 seconds = 22.7 seconds, increasing the interval to 7.4 seconds. The predicted information entropy was 5.19, a larger decrease.

[0149] The fitness evaluation metric was calculated after each adjustment. For the second exposure time point adjusted to 15.3 seconds, the new mean square error (MSE) was calculated to be 0.0075, a slight improvement over the original 0.0078. For the third exposure time point adjusted to 22.2 seconds, the new MSE was calculated to be 0.0082, a slight increase but still within an acceptable range. For the third exposure time point adjusted to 22.7 seconds, the new MSE was calculated to be 0.0097, a significant increase.

[0150] The time offset step size is iteratively optimized using the gradient descent method. In this example, the iterative process of the gradient descent method is as follows: First, the gradient of the fitness with respect to the time offset is calculated, that is, the rate of change of the fitness with the time offset. For the second exposure time point, the gradient of the fitness with time offset is -0.0006 / 0.5 = -0.0012, indicating that the fitness decreases by 0.0006 for every 0.5 second increase. For the third exposure time point, the gradient of the fitness with time offset is 0.0004 / 0.5 = 0.0008 (0.5 second offset), 0.0008 / 0.5 = 0.0016 (1.0 second offset), and 0.0030 / 0.5 = 0.0060 (1.5 second offset), showing a gradually increasing trend.

[0151] Based on the gradient calculation results, the optimal time offset step size was determined. For the second exposure time point, since the gradient was negative, a positive offset was appropriate, and a 0.5 second offset step size was ultimately selected, resulting in an adjustment of 15.3 seconds. For the third exposure time point, considering that the gradient increases with offset, and that the gradients for 0.5 and 1.0 second offsets are smaller, a 1.0 second offset step size was ultimately selected, resulting in an adjustment of 22.2 seconds.

[0152] Similar adjustments were performed on all exposure time points with a presence status of 1, resulting in the optimized predicted exposure time series of 7.5, 15.3, 22.2, 30.2, 42.0, and 57.8 seconds. The intervals between adjacent exposure time points were 7.8, 6.9, 8.0, 11.8, and 15.8 seconds, respectively, all meeting the minimum time interval threshold requirement. The optimized predicted information entropies were 3.92, 5.89, 5.26, 4.43, 2.94, and 1.72, respectively.

[0153] The fitness evaluation metric for the optimized exposure time series prediction scheme was calculated, and the mean square error (MSE) was 0.0069, a significant improvement over the original 0.0078. This indicates that by properly adjusting the exposure time points, not only did the time interval requirements meet, but the prediction accuracy was also improved.

[0154] The final optimized predicted exposure time series scheme was applied to the patient's DSA examination. The information entropy of the actual image acquisition was 3.98, 5.94, 5.31, 4.39, 2.89, and 1.68, respectively. The average error from the predicted value was only 1.7%, verifying the effectiveness of the optimization method.

[0155] Example 2: Based on the same inventive concept, Figure 2 As shown, this embodiment also provides a low-latency brain electrophysiological monitoring system during DSA surgery, the system comprising:

[0156] An exposure image acquisition module is used to acquire continuous multiple-frame scan images after contrast agent injection during DSA surgery using a non-developing electrode, and to establish an initial diffusion image sequence including the scan time and the corresponding image data; the image information entropy is obtained by the grayscale distribution of each frame of the image, and a mapping relationship between the scan time and the image information entropy in the initial diffusion image sequence is established;

[0157] The parameter iteration initialization module is used to generate the initial population of the predicted exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes the existence state, the predicted time value, and the predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy.

[0158] A parameter iteration module is used to iteratively evolve the initial population. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with a state of 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold.

[0159] The dynamic exposure time adjustment module is used to update the mapping relationship between the scanning time and the image information entropy after each scan is completed; dynamically adjust the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence according to the updated mapping relationship; set the minimum time interval threshold between adjacent exposure time points, and eliminate exposure time points with a time interval less than the threshold; and use the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached as the subsequent exposure time.

[0160] Furthermore, the system further comprises:

[0161] X-ray image information entropy calculation module is used to obtain the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area ;No. Contrast agent diffusion information entropy at each scanning moment The calculation method is:

[0162] ;

[0163] in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent;

[0164] No. Vascular imaging quality information entropy at each scanning moment The calculation method is:

[0165] ;

[0166] in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

[0167] Furthermore, the system further comprises:

[0168] Contrast agent diffusion modeling module, used to obtain the first Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ;

[0169] The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within :

[0170] ;

[0171] in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representative sub-region The number of pixels included;

[0172] Calculate the Contrast agent diffusion intensity at each scanning moment :

[0173] ;

[0174] in Representative The average gray value of all sub-areas at a scanning moment:

[0175] ;

[0176] Calculate the Contrast agent diffusion area at each scanning moment :

[0177] .

[0178] Furthermore, the system further comprises:

[0179] Genetic iteration module is used to convert the first Individual Represented as a set of forecast exposure time series schemes:

[0180] ;

[0181] in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable;

[0182] No. Predicting exposure time series schemes Fitness evaluation index The calculation method is:

[0183] ;

[0184] in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1;

[0185] According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for:

[0186] ;

[0187] in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

[0188] Furthermore, the system further comprises:

[0189] Exposure moment fine-tuning module, used to obtain the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ;

[0190] The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

[0191] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0192] Finally, it should be noted that although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A low-latency electrophysiological monitoring method for brain surgery during DSA surgery, characterized in that: The method comprises: Acquire multiple consecutive scanned images after contrast agent injection during DSA surgery using a non-developing electrode, and establish an initial diffusion image sequence including the scan time and corresponding image data; obtain image information entropy through the grayscale distribution of each frame image, and establish a mapping relationship between the scan time and image information entropy in the initial diffusion image sequence; Generate an initial population for predicting exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes an existence state, a predicted time value, and a predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy. The initial population is iteratively evolved. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with state 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold. After each scan is completed, the mapping relationship between the scan time and the image information entropy is updated; the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence is dynamically adjusted based on the updated mapping relationship; a minimum time interval threshold between adjacent exposure time points is set, and exposure time points with a time interval less than the threshold are eliminated; and the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached are used as subsequent exposure moments; The method of acquiring a plurality of consecutive scanned images after contrast agent injection in a DSA surgery using a non-developing electrode, establishing an initial diffusion image sequence including scanning time and corresponding image data; obtaining image information entropy through the grayscale distribution of each frame of image, and establishing a mapping relationship between scanning time and image information entropy in the initial diffusion image sequence includes: Get the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extract the contrast agent diffusion intensity and diffusion area in the brain vascular region through angiographic difference images; Contrast agent diffusion information entropy at each scanning moment The calculation method is: ； in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent; No. Vascular imaging quality information entropy at each scanning moment The calculation method is: ； in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

2. The low-latency electrophysiological monitoring method for brain surgery during DSA surgery according to claim 1, characterized in that: The first image in the initial diffusion image sequence is obtained The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extracting contrast agent diffusion intensity in brain vascular regions using angiographic difference images and diffusion area The methods include: Get the Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ; The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within : ； in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representing sub-regions The number of pixels included; Calculate the Contrast agent diffusion intensity at each scanning moment : ； in Representative The average gray value of all sub-areas at a scanning moment: ； Calculate the Contrast agent diffusion area at each scanning moment : 。 3. The low-latency electrophysiological brain monitoring method during DSA surgery according to claim 2, characterized in that: The initial population is iteratively evolved, wherein the iterative evolution step includes calculating the mean square difference between the predicted information entropy of the exposure time point with a state of 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation index, and the method of selecting a preset number of winning individuals through a roulette wheel selection method based on the fitness index includes: The first Individual Represented as a set of forecast exposure time series schemes: ； in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable; No. Predicting exposure time series schemes Fitness evaluation index The calculation method is: ； in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1; According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for: ； in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

4. The low-latency electrophysiological monitoring method for brain surgery during DSA surgery according to claim 3, characterized in that: The method further comprises: Get the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ; The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

5. A low-latency brain electrophysiological monitoring system during DSA surgery, characterized in that: The system comprises: An exposure image acquisition module is used to acquire continuous multi-frame scan images after contrast agent injection during DSA surgery using a non-developing electrode, and to establish an initial diffusion image sequence including the scan time and the corresponding image data; the image information entropy is obtained by the grayscale distribution of each frame of the image, and a mapping relationship between the scan time and the image information entropy in the initial diffusion image sequence is established; The parameter iteration initialization module is used to generate the initial population of the predicted exposure time series. The initial population includes multiple individuals, each of which represents a predicted exposure time series scheme. Each predicted exposure time series scheme includes multiple exposure time points. The parameter state of each exposure time point includes the existence state, the predicted time value, and the predicted information entropy. The existence state takes a value of 0 or 1, and the predicted time value is randomly generated within the time range of the initial diffusion image sequence. The predicted information entropy is obtained by mapping the scanning time and the image information entropy. A parameter iteration module is used to iteratively evolve the initial population. The iterative evolution steps include calculating the mean square difference between the predicted information entropy of the exposure time point with a state of 1 in each predicted exposure time series scheme and the information entropy of the actual collected image as a fitness evaluation indicator. A preset number of winning individuals are selected through a roulette wheel selection method based on the fitness indicator. The winning individuals are crossover and mutated to obtain offspring individuals. The initial population is replaced with the offspring individuals, and the iterative evolution process is repeated until the fitness improvement value of the offspring individuals and the parent individuals is lower than the set fitness threshold. The dynamic exposure time adjustment module is used to update the mapping relationship between the scanning time and the image information entropy after each scan is completed; dynamically adjust the predicted information entropy of the exposure time points to be collected in the predicted exposure time sequence based on the updated mapping relationship; set the minimum time interval threshold between adjacent exposure time points, and eliminate exposure time points with a time interval less than the threshold; and use the time points in the winning individuals after the iterative evolution with a state of 1 and a time value not yet reached as the subsequent exposure time; The system further comprises: X-ray image information entropy calculation module is used to obtain the first The image corresponding to the scanning moment, the image Subtract the mask image before contrast agent injection Obtain angiographic difference images ; Extract the contrast agent diffusion intensity and diffusion area in the brain vascular region through angiographic difference images; Contrast agent diffusion information entropy at each scanning moment The calculation method is: ； in For the The average gray value of the contrast agent in the blood vessel area at each scanning moment, is the maximum average gray value observed during the diffusion of contrast agent; For the The number of pixels of the contrast agent diffusion area in the blood vessel area at each scanning moment, is the number of pixels of the maximum diffusion area observed during the diffusion of contrast agent; No. Vascular imaging quality information entropy at each scanning moment The calculation method is: ； in For the The scanning time is The gray gradient value of the blood vessel branch area, For the The sum of the grayscale gradient values ​​of all blood vessel branch areas at each scanning moment, is the total number of vascular branch areas; the scanning time Information entropy of contrast agent diffusion and vascular imaging quality information entropy The product of Create a mapping function , get any time by cubic spline interpolation Image information entropy prediction value .

6. The low-latency electrophysiological brain monitoring system during DSA surgery according to claim 5, characterized in that: The system further comprises: Contrast agent diffusion modeling module, used to obtain the first Angiographic difference images at each scanning moment , the angiographic difference image Perform binarization to obtain a binary mask of the brain vascular area ; Use morphological operations to perform binary mask Boundary optimization, including dilation and erosion operations, is performed to obtain the optimized brain vascular area mask. ; The optimized brain vascular area mask Divided into sub-regions , where each sub-region The boundary of is determined by the region growing algorithm. The value ranges from 1 to Integer variable; calculate each sub-region The average gray value within : ； in represents pixel coordinates, Represents angiographic difference images at coordinates The gray value at Representing sub-regions The number of pixels included; Calculate the Contrast agent diffusion intensity at each scanning moment : ； in Representative The average gray value of all sub-areas at a scanning moment: ； Calculate the Contrast agent diffusion area at each scanning moment : 。 7. The low-latency electrophysiological brain monitoring system during DSA surgery according to claim 6, characterized in that: The system further comprises: Genetic iteration module is used to convert the first Individual Represented as a set of forecast exposure time series schemes: ； in Representative Predicting exposure time series schemes No. The existence status of each exposure time point, the value is 0 or 1; Representative Predicting exposure time series schemes No. The predicted time value of each exposure time point; Representative Predicting exposure time series schemes No. The predicted information entropy of each exposure time point, is a positive integer variable, The value ranges from 1 to the number of exposure time points integer variable; No. Predicting exposure time series schemes Fitness evaluation index The calculation method is: ； in Representative Predicting exposure time series schemes No. The information entropy of the actual collected image corresponding to each exposure time point is: Representative Predicting exposure time series schemes Existence state The number of exposure time points is 1; According to the calculated fitness evaluation index , sort the predicted exposure time series schemes by fitness, and select a preset number of winning individuals through the roulette wheel selection method, where the first Predicting exposure time series schemes Probability of being selected for: ； in Represents the maximum fitness evaluation index value among all prediction exposure time series schemes, The total number of forecast exposure time series scenarios representing the initial population.

8. The low-latency electrophysiological brain monitoring system during DSA surgery according to claim 7, characterized in that: The system further comprises: Exposure moment fine-tuning module, used to obtain the predicted exposure time series solution Existence state For exposure time points with a value of 1, the minimum time interval threshold between adjacent exposure time points is set as the time offset threshold ; Set the time offset threshold Split into Time migration step size is obtained by dividing , for each state of existence The predicted time value is 1 Adjust in sequence; the adjusted forecast time value ,in The value ranges from 1 to Integer variable; adjusted forecast time value The time interval between the exposure time point with the adjacent existence state 1 must be no less than the time offset threshold ; The adjusted forecast time value Substitute into the mapping function Get the predicted information entropy , calculate the adjusted fitness evaluation index , the time offset step is adjusted by gradient descent method Perform iterative optimization and use the time offset step obtained by the final iteration As the predicted time value The correction parameters are used to update the corresponding predicted time value and predicted information entropy in the predicted exposure time series scheme to obtain the final optimized predicted exposure time series scheme.

Images