Medical image instant feedback system based on edge calculation

Abstract

The invention relates to the technical field of medical images, in particular to a medical image instant feedback system based on edge computing, which comprises an edge performance monitoring module, an emergency image scheduling module, a real-time image analysis module and a feedback efficiency optimization module, and continuously monitoring node performance, updating edge node performance indexes in real time, comparing the edge node performance indexes with historical data, and marking according to deviation data and time. According to the invention, through continuous tracking of node performance and real-time updating of a processing strategy, the response capability of the system to an acute disease image is improved, through application of DQN and CNN, the recognition efficiency of feature extraction and emergency signs is enhanced, the self-optimization capability of the system is improved through adaptive adjustment, and the adaptability to changed medical data features is also enhanced. The method is crucial to coping with medical image requirements which change rapidly, so that the quality of medical services is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to an edge computing-based medical imaging instant feedback system. Background Art

[0002] Medical imaging technology is a technology used in clinical diagnosis and medical research. It involves the generation, processing and analysis of images of the inside of the body, including multiple imaging methods such as X-ray imaging, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound imaging and positron emission tomography (PET). The development of medical imaging technology has greatly improved the accuracy of disease diagnosis and enabled doctors to view the internal structure of the body without invasive surgery.

[0003] Among them, the medical imaging instant feedback system is a system that uses medical imaging technology to provide real-time analysis. Through computing technology, the system can quickly process and analyze medical imaging data, thereby providing instant feedback during the patient examination process. Its main purpose is to improve the speed and accuracy of diagnosis and reduce the possibility of misdiagnosis. In emergency medical situations, such as heart attacks or serious trauma, instant feedback is particularly critical and can help make quick decisions and treat patients.

[0004] In terms of data processing speed and real-time performance, existing systems rely on central processing units to process data, which often leads to slow response speeds in high data flow environments. This is particularly prominent in emergency medical imaging scenarios that require rapid feedback, such as delays in processing heart disease or severe trauma images, and missed critical opportunities to provide timely feedback to patients. When responding to rapidly changing medical imaging needs, existing systems often affect the accuracy and timeliness of image analysis due to lagging strategy updates, affecting the application effect of instant feedback on medical images. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art and propose a medical image instant feedback system based on edge computing.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a medical image instant feedback system based on edge computing, the system comprising:

[0007] The edge performance monitoring module continuously monitors the node performance based on the processing speed and activity data of the edge node, updates the edge node performance indicators in real time, and compares them with historical data. According to the deviation data and time stamp, it adjusts the data collection frequency and generates monitoring parameter adjustment information;

[0008] The emergency imaging scheduling module uses DQN to identify high-priority imaging tasks including acute diseases based on the monitoring parameter adjustment information, recalculates task allocation according to the real-time performance of edge nodes, optimizes the task processing sequence, and generates a task priority ranking table;

[0009] The real-time image analysis module uses CNN to analyze image data based on the task priority ranking table, extracts key medical features and emergency signs, and compares and analyzes the extracted features with historical feedback data to evaluate the matching degree and generate key feature matching results;

[0010] The feedback efficiency optimization module calculates feedback accuracy and deviation based on the key feature matching results and the current feedback results, refines and adjusts the parameter settings and priority queues of the data flow analysis process, and generates tuned feedback parameters.

[0011] The present invention is improved in that the real-time update step of the edge node performance indicator is specifically as follows:

[0012] Based on the processing speed and activity data of the edge nodes, collect the processing speed and activity, and for each edge node, use the formula:

[0013]

[0014] Calculate the performance index P t , where n represents the total number of data points, v i represents the processing speed of the i-th node, d i represents the activity level, k v and k d are the corresponding weight coefficients respectively;

[0015] Based on the performance index, the deviation is calculated by comparing with the historical performance index, using the formula:

[0016] ΔP=P t -P h

[0017] The performance deviation ΔP is obtained, where P t is the performance index, P h is the historical performance index;

[0018] Based on the performance deviation, the node performance index is updated using the formula:

[0019] P new =P t -γ·ΔP

[0020] Get the updated edge node performance index P new , where P tis the performance index, ΔP is the performance deviation, and γ is the performance adjustment factor.

[0021] The present invention is improved in that the step of obtaining the monitoring parameter adjustment information is specifically as follows:

[0022] Based on the deviation data, combined with the time stamp, the formula is used:

[0023]

[0024] Calculate the new data collection frequency f new , where f old is the original collection frequency, ΔP is the performance deviation, P threshold is the performance deviation threshold, k q is the adjustment coefficient, t is the time factor;

[0025] Analyze the difference between the new data collection frequency and the original frequency, and use the judgment formula:

[0026] AD=if(|f new -f old |>∈ q )then "update"else "keep"

[0027] Generate decision result AD and obtain monitoring parameter adjustment information, where f new is the new data collection frequency, f old is the original collection frequency, ∈ q is the sensitivity threshold.

[0028] The present invention is improved in that the identification step of the high priority image task is specifically as follows:

[0029] Based on the monitoring parameter adjustment information, tasks including acute conditions are screened from the medical imaging dataset, analyzed by the DQN model, and the urgency is determined using the formula:

[0030] U i =β1·AI score +β2·Clinical urgency

[0031] Get the urgency score U of task i i , among which, AI score Score for AI analysis, Clinical urgency is the clinical urgency score, β1 and β2 are the weight coefficients of the adjustment factors;

[0032] The urgency scores are ranked using the formula:

[0033] U sorted = sort(Ui , "Descending")

[0034] Get the sorted task list U sorted , and select key priority tasks according to the task list to obtain a high priority list, where U i is the urgency score of task i.

[0035] The present invention is improved in that the optimization steps of the task processing sequence are specifically as follows:

[0036] According to the real-time performance of the edge node and the current task load, the formula is adopted:

[0037]

[0038] Calculate new processing power EC new , among which, EP current ET is the real-time performance index. load is the number of tasks on the node;

[0039] Based on the new processing capacity, the task priority is recalculated according to the urgency score, using the formula:

[0040]

[0041] Get the new priority EP of the task task , where U i is the task urgency score, EC new It’s new processing power;

[0042] According to the new priorities of the tasks, the tasks are sorted in descending order again, the task processing sequence is optimized, and a task priority sorting table is obtained.

[0043] The present invention is improved in that the steps of extracting the medical features and emergency signs are specifically as follows:

[0044] Based on the task priority ranking table, the input medical imaging data is analyzed using a convolutional neural network using the formula:

[0045]

[0046] Extract the medical feature set F extracted ,in, represents the i-th segment in the input medical image data, WR i is the weight matrix in the CNN layer, b R is the bias term, n R is the number of features or filters in the convolutional layer;

[0047] Based on the medical feature set, identification is performed through an emergency threshold, and if the feature set exceeds the emergency threshold, it is determined to be an emergency sign;

[0048] The medical feature set and emergency signs are integrated using the formula:

[0049] F integrated =F extracted ·U flags

[0050] Get integrated medical features and marker data F integrated , where U flags It is an emergency sign. extracted is the extracted medical feature set.

[0051] The present invention is improved in that the step of obtaining the key feature matching result is specifically as follows:

[0052] Based on the medical characteristics and emergency signs, compared with historical medical data, the formula is used:

[0053]

[0054] Generate a difference comparison result RY, where F integrated [i] is the i-th integrated medical feature and marker data, H data [i] is the corresponding ith eigenvalue in the historical data, k Y is the number of features;

[0055] Based on the difference comparison results, the matching degree is evaluated using the formula:

[0056] MY=exp(-α Y RY)

[0057] Get the matching degree MY, where RY is the difference comparison result, α Y is a tuning parameter used to control the matching sensitivity.

[0058] The present invention is improved in that the step of obtaining the feedback parameter for tuning is specifically as follows:

[0059] Based on the key feature matching results, the error vector is calculated according to the current measurement value and the feedback result, using the formula:

[0060]

[0061] Get the accuracy evaluation result SD, where FH current [i] is the i-th eigenvalue of the current feedback, OF actual [i] is the i-th eigenvalue of the current observation, UO iis the weight of the i-th feature, pJ is the norm order, n V is the total number of features;

[0062] Based on the accuracy evaluation results, the parameters are adjusted, and the degree of parameter update is adjusted using the hyperbolic tangent function, using the formula:

[0063] ZP new =ZP old ·(1+β Z tanh(γ Z SD-δ Z ))

[0064] Get new parameter settings ZP new , where ZP old is the original parameter setting, β Z , γ Z and δ Z is the coefficient for adjusting sensitivity, scale, and bias, and SD is the accuracy assessment result;

[0065] Apply the new parameter settings, optimize the data processing flow and priority queue, and update the configuration based on the feedback to obtain the tuned feedback parameters.

[0066] Compared with the prior art, the advantages and positive effects of the present invention are:

[0067] In the present invention, the system's responsiveness to acute disease images is improved by continuously tracking node performance and updating processing strategies in real time. The application of DQN and CNN enhances the efficiency of feature extraction and recognition of emergency signs. Adaptive adjustment improves the system's self-optimization capability and also enhances its adaptability to changing medical data features, which is crucial for responding to rapidly changing medical imaging needs, thereby improving the quality of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a system flow chart of the present invention;

[0069] Figure 2 A real-time update flow chart of edge node performance indicators in the present invention;

[0070] Figure 3 A flowchart for obtaining monitoring parameter adjustment information in the present invention;

[0071] Figure 4 is a flow chart for identifying high priority image tasks in the present invention;

[0072] Figure 5 It is an optimized flow chart of the task processing sequence in the present invention;

[0073] Figure 6is a flowchart of extracting traditional Chinese medicine characteristics and emergency signs of the present invention;

[0074] Figure 7 This is a flowchart for obtaining key feature matching results in the present invention;

[0075] Figure 8 This is a flow chart for obtaining the optimized feedback parameters in the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0077] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0078] Example

[0079] See also Figure 1 The present invention provides a technical solution: a medical image instant feedback system based on edge computing includes:

[0080] The edge performance monitoring module continuously monitors the node performance based on the processing speed and activity data of the edge node, updates the edge node performance indicators in real time, and compares them with historical data. According to the deviation data and time stamp, it adjusts the data collection frequency and generates monitoring parameter adjustment information;

[0081] The emergency imaging scheduling module uses DQN to identify high-priority imaging tasks involving acute diseases based on monitoring parameter adjustment information, recalculates task allocation according to the real-time performance of edge nodes, optimizes task processing sequences, and generates a task priority ranking table;

[0082] The real-time image analysis module uses CNN to analyze image data based on the task priority table, extract key medical features and emergency signs, and compares and analyzes the extracted features with historical feedback data to evaluate the matching degree and generate key feature matching results;

[0083] The feedback efficiency optimization module calculates the feedback accuracy and deviation based on the key feature matching results and the current feedback results, refines and adjusts the parameter settings and priority queues of the data flow analysis process, and generates tuned feedback parameters.

[0084] The monitoring parameter adjustment information includes node performance differences, data collection intervals, and historical performance comparison indicators. The task priority ranking table includes the urgency level, task response time, and resource requirements. The key feature matching results include feature recognition accuracy, emergency sign verification results, and matching rating results. The tuned feedback parameters include feedback cycle, accuracy improvement indicators, and deviation correction process.

[0085] See also Figure 2 ,The real-time update steps of the edge node performance indicators are as follows:

[0086] Based on the processing speed and activity data of the edge nodes, collect the processing speed and activity, and for each edge node, use the formula:

[0087]

[0088] Calculate the performance index P t , where n represents the total number of data points, v i represents the processing speed of the i-th node, d i represents the activity level, k v and k d are the corresponding weight coefficients respectively;

[0089] Based on the performance index, the deviation is calculated by comparing it with the historical performance index, using the formula:

[0090] ΔP=P t -P h

[0091] The performance deviation ΔP is obtained, where P t is the performance index, P h is the historical performance index;

[0092] Based on the performance deviation, update the node performance index using the formula:

[0093] P new =P t -γ·ΔP

[0094] Get the updated edge node performance index P new , where P t is the performance index, ΔP is the performance deviation, and γ is the performance adjustment factor.

[0095] k v and k d Set to 0.7 and 0.3;

[0096] Calculation process:

[0097] The processing speeds of the 10 nodes are collected as follows:

[0098] {20, 25, 30, 35, 20, 25, 30, 35, 40, 45} packets / second;

[0099] The activities are:

[0100] {50, 45, 40, 35, 50, 45, 40, 35, 30, 25} packets / sec;

[0101] Calculation formula:

[0102]

[0103] The contribution of the first node is 0.7×20+0.3×50=14+15=29 packets / second;

[0104] The obtained P t The value represents the weighted average node performance index, which is used to evaluate the real-time processing capability of the entire system.

[0105] Detection of P h 30 packets / second;

[0106] Calculation process:

[0107] ΔP=32-30=2

[0108] ΔP is 2 packets / second, which indicates the difference between current performance and historical performance and is used to determine whether performance has changed significantly.

[0109] γ is set to 0.1;

[0110] Using ΔP = 2 packets / second and P t =32 packets / second;

[0111] Calculation process:

[0112] P new =32-0.1×2=32-0.2=31.8

[0113] P new It is 31.8 packets / second, which represents the new adjusted performance index, providing an up-to-date assessment of the node’s current performance, taking into account recent performance changes.

[0114] See also Figure 3 , the specific steps for obtaining the monitoring parameter adjustment information are:

[0115] Based on the deviation data, combined with the time stamp, use the formula:

[0116]

[0117] Calculate the new data collection frequency f new , where f old is the original collection frequency, ΔP is the performance deviation, P threshold is the performance deviation threshold, k q is the adjustment coefficient, t is the time factor;

[0118] Analyze the difference between the new data collection frequency and the original frequency, and use the judgment formula:

[0119] AD=if(|f new -f old |>∈ q )then "update"else "keep"

[0120] Generate decision result AD and obtain monitoring parameter adjustment information, where f new is the new data collection frequency, f old is the original collection frequency, ∈ q is the sensitivity threshold.

[0121] The following data is collected:

[0122] f old 1Hz;

[0123] ΔP is 0.05;

[0124] P threshold Set to 0.1;

[0125] k q Set to 0.5;

[0126] t is 2.

[0127] Calculation process:

[0128]

[0129] f new =1×(1+0.5·0.5) 2

[0130] f new =1×(1+0.25) 2

[0131] f new =1×1.25 2

[0132] f new=1×1.5625=1.5625Hz

[0133] New data collection frequency f new is 1.5625 Hz, which indicates that according to the adjustment of performance deviation and time factor, the data collection frequency should be increased to monitor the performance changes more intensively.

[0134] Setting ∈ q 0.1Hz;

[0135] |f new -f old |=|1.5625-1|=0.5625

[0136] Then 0.5625 is greater than 0.1, and the decision result should be update.

[0137] See also Figure 4 , the recognition steps of high-priority image tasks are as follows:

[0138] Based on the monitoring parameter adjustment information, tasks involving acute conditions are screened from the medical imaging dataset, analyzed through the DQN model, and the urgency is determined using the formula:

[0139] U i =β1·AI score +β2·Clinical urgency

[0140] Get the urgency score U of task i i , among which, AI score Score for AI analysis, Clinical urgency is the clinical urgency score, β1 and β2 are the weight coefficients of the adjustment factors;

[0141] To sort the urgency scores, use the formula:

[0142] U sorted = sort(U i , "Descending")

[0143] Get the sorted task list U sorted , and select key priority tasks according to the task list to obtain a high priority list, where U i is the urgency score of task i.

[0144] Set β1=0.6 and β2=0.4.

[0145] When the AI ​​for a task score =0.85 and Clinical urgency =0.70, calculate:

[0146] U i =0.6·0.85+0.4·0.70

[0147] U i =0.51+0.28=0.79

[0148] Based on AI analysis and clinical assessment, the urgency score of task i was 0.79, and the score was used for further task prioritization.

[0149] U has four tasks i The scores were 0.79, 0.55, 0.83 and 0.65 respectively.

[0150] U sorted =sort([0.79,0.55,0.83,0.65], "descending")

[0151] U sorted =[0.83,0.79,0.65,0.55]

[0152] The sorted task list shows that Task 3 (0.83) has the highest urgency score.

[0153] See also Figure 5 , the optimization steps of the task processing sequence are as follows:

[0154] According to the real-time performance of the edge node and the current task load, the formula is used:

[0155]

[0156] Calculate new processing power EC new , among which, EP current ET is the real-time performance index. load is the number of tasks on the node;

[0157] Based on the new processing capacity, the task priority is recalculated according to the urgency score, using the formula:

[0158]

[0159] Get the new priority EP of the task task , where U i is the task urgency score, EC new It’s new processing power;

[0160] According to the new priority of the task, the tasks are sorted in descending order again, the task processing sequence is optimized, and a task priority sorting table is obtained to ensure that urgent tasks are executed first.

[0161] The following data was collected:

[0162] EPcurrent is 100 tasks / second;

[0163] ET load For 20 tasks;

[0164] Substituting the values ​​into the formula:

[0165]

[0166] Get EC new It is 4.76 tasks / second, which means that under the current load, the node can process about 4.76 tasks per second, which is the new processing capacity of the node after adjustment.

[0167] Collection of U i is 0.85;

[0168] Substituting the values ​​into the formula:

[0169]

[0170] EP task A value of 0.175 indicates that the priority score of the task is 0.175 after considering the current processing capacity. A lower value indicates that even if the task is urgent, its priority is relatively lower due to the high load of the node.

[0171] See also Figure 6 ,The steps for extracting medical features and emergency signs are as follows:

[0172] Based on the task priority table, the convolutional neural network is used to analyze the input medical imaging data using the formula:

[0173]

[0174] Extract the medical feature set F extracted ,in, represents the i-th segment in the input medical image data, WR i is the weight matrix in the CNN layer, b R is the bias term, n R is the number of features or filters in the convolutional layer;

[0175] Based on the medical feature set, the emergency threshold is used for identification. If the feature set exceeds the emergency threshold, it is judged as an emergency sign;

[0176] The medical feature set and emergency signs are integrated using the formula:

[0177] F integrated =F extracted ·U flags

[0178] Get integrated medical features and marker data F integrated , where U flags It is an emergency sign. extracted is the extracted medical feature set.

[0179] Assume that CNN has only one filter, weight WR1 = [0.2, -0.1], bias b R =0.05, input

[0180] data

[0181] F extracted =[0.2,-0.1]·[0.9,0.8]+0.05

[0182] F extracted =0.2×0.9+(-0.1)×0.8+0.05

[0183] F extracted =0.18-0.08+0.05=0.15

[0184] It means that for the given image data and CNN model configuration, the extracted feature value is 0.15 for emergency sign detection.

[0185] If the extracted feature value is 0.15, and the emergency threshold is set to 0.1, it is judged that 0.15>0.1, so the extracted feature is classified as an emergency sign;

[0186] F integrated =0.15·1=0.15

[0187] Indicates F integrated Keep it at 0.15 and include information about emergency flags.

[0188] See also Figure 7 , the specific steps for obtaining key feature matching results are:

[0189] Based on medical characteristics and emergency signs, compared with historical medical data, the formula is used:

[0190]

[0191] Generate a difference comparison result RY, where F integrated [i] is the i-th integrated medical feature and marker data, H data [i] is the corresponding ith eigenvalue in the historical data, k Y is the number of features;

[0192] Based on the difference comparison results, the matching degree is evaluated using the formula:

[0193] MY=exp(-α Y RY)

[0194] The matching degree MY is obtained, where RY is the difference comparison result and αY is an adjustment parameter used to control the matching sensitivity.

[0195] Two features are collected and monitored:

[0196] F integrated =[0.9,0.8]

[0197] H data =[0.5,0.6]

[0198] k Y =2

[0199] Calculation process:

[0200]

[0201] The RY value is 0.447, which represents the Euclidean distance between two sets of data and is used to assess the degree of difference between features.

[0202] Assume α Y =0.1.

[0203] Calculation process:

[0204] MY=exp(-0.1×0.447)

[0205] MY=exp(-0.0447)

[0206] MY=0.956

[0207] The MY value is 0.956, which is close to 1, indicating a high match, meaning that the current features are highly similar to the historical data, pointing to the same or similar medical conditions.

[0208] See also Figure 8 , the specific steps for obtaining the feedback parameters for tuning are:

[0209] Based on the key feature matching results, the error vector is calculated according to the current measurement value and feedback results, using the formula:

[0210]

[0211] Get the accuracy evaluation result SD, where FH current [i] is the i-th eigenvalue of the current feedback, OF actual [i] is the i-th eigenvalue of the current observation, UO i is the weight of the i-th feature, pJ is the norm order, n Vis the total number of features;

[0212] Based on the accuracy evaluation results, the parameters are adjusted and the hyperbolic tangent function is used to adjust the parameter update degree. The formula is:

[0213] ZP new =ZP old ·(1+β Z tanh(γ Z SD-δ Z ))

[0214] Get new parameter settings ZP new , where ZP old is the original parameter setting, β Z , γ Z and δ Z is the coefficient for adjusting sensitivity, scale, and bias, and SD is the accuracy assessment result;

[0215] Apply new parameter settings, optimize data processing flow and priority queues, and update configurations based on feedback to obtain tuned feedback parameters.

[0216] There are three characteristics to collect feedback and actual values:

[0217] FH current =[100,150,200];

[0218] OF actual =[110,140,210];

[0219] UO = [0.5, 1, 0.5];

[0220] pJ=2.

[0221] Calculation process:

[0222]

[0223] SD = 12.25 indicates the overall deviation between the current feedback and the actual observed data.

[0224] Set up and collect the following data:

[0225] ZP old =0.8;

[0226] β Z =0.1;

[0227] γ Z =1;

[0228] δ Z =5;

[0229] SD=12.25.

[0230] Calculation process:

[0231] ZP new =0.8·(1+0.1·tanh(1·12.25-5))

[0232] ZP new =0.8·(1+0.1·tanh(7.25))

[0233] ZP new =0.8·1.07=0.856

[0234] ZP new = 0.856 reflects the error-adjusted parameter settings aimed at optimizing system performance.

[0235] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A medical image instant feedback system based on edge computing, characterized in that: The system comprises: The edge performance monitoring module continuously monitors node performance based on edge node processing speed and activity data, updates edge node performance indicators in real time, and compares them with historical data. Based on deviation data and time stamps, it adjusts data collection frequency and generates monitoring parameter adjustment information. The emergency imaging scheduling module uses DQN to identify high-priority imaging tasks including acute conditions based on the monitoring parameter adjustment information, recalculates task allocation according to the real-time performance of edge nodes, optimizes the task processing sequence, and generates a task priority ranking table; The real-time image analysis module uses CNN to analyze image data based on the task priority ranking table, extracts key medical features and emergency signs, compares and analyzes the extracted features with historical feedback data, evaluates the degree of matching, and generates key feature matching results; The feedback efficiency optimization module calculates the feedback accuracy and deviation based on the key feature matching results and the current feedback results, refines and adjusts the parameter settings and priority queues of the data flow analysis process, and generates tuned feedback parameters.

2. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The real-time update steps of the edge node performance indicators are specifically as follows: Based on the processing speed and activity data of the edge nodes, the processing speed and activity amount are collected. For each edge node, the formula is used: Calculate the performance index P t , where n represents the total number of data points, v i represents the processing speed of the i-th node, d i represents the activity level, k v and k d are the corresponding weight coefficients respectively; Based on the performance index, the deviation is calculated by comparing it with the historical performance index using the formula: ΔP=P t -P h The performance deviation ΔP is obtained, where P t is the performance index, P h is the historical performance index; Based on the performance deviation, the node performance index is updated using the formula: P new =P t -γ·ΔP Get the updated edge node performance index P new , where P t is the performance index, ΔP is the performance deviation, and γ is the performance adjustment coefficient.

3. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The steps for obtaining the monitoring parameter adjustment information are specifically as follows: Based on the deviation data, combined with the time stamp, the formula is used: Calculate the new data collection frequency f new , where f old is the original collection frequency, ΔP is the performance deviation, P threshold is the performance deviation threshold, k q is the adjustment coefficient, t is the time factor; Analyze the difference between the new data collection frequency and the original frequency and use the judgment formula: AD=if(|f new -f old |>∈ q )then "update"else "keep" Generate decision result AD and obtain monitoring parameter adjustment information, where f new is the new data collection frequency, f old is the original collection frequency, ∈ q is the sensitivity threshold.

4. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The identification steps of the high-priority image task are specifically as follows: Based on the monitoring parameter adjustment information, tasks including acute conditions are screened from the medical imaging dataset, analyzed using the DQN model, and the urgency is determined using the formula: U i =β1·AI score +β2·Clinical urgency Get the urgency score U of task i i , among which AI score Score for AI analysis, Clinical urgency is the clinical urgency score, β1 and β2 are the weight coefficients of the adjustment factors; The urgency scores are ranked using the formula: U sorted =sort(U i , "Descending") Get the sorted task list U sorted , and select key priority tasks according to the task list to obtain a high priority list, where U i is the urgency score of task i.

5. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The optimization steps of the task processing sequence are specifically as follows: According to the real-time performance of the edge node and the current task load, the formula is adopted: Calculate new processing power EC new , among which, EP current ET is the real-time performance index. load is the number of tasks on the node; Based on the new processing capacity, the task priority is recalculated according to the urgency score, using the formula: Get the new priority EP of the task task , where U i is the task urgency score, EC new It’s new processing power; According to the new priorities of the tasks, the tasks are sorted in descending order again, the task processing sequence is optimized, and a task priority sorting table is obtained.

6. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The steps for extracting the medical features and emergency signs are specifically as follows: Based on the task priority table, a convolutional neural network is used to analyze the input medical imaging data using the formula: Extract the medical feature set F extracted ,in, Represents the i-th segment in the input medical image data, WR i is the weight matrix in the CNN layer, b R is the bias term, n R is the number of features or filters in the convolutional layer; Based on the medical feature set, identification is performed using an emergency threshold, and if the feature set exceeds the emergency threshold, it is determined to be an emergency sign; The medical feature set and emergency sign are integrated using the formula: F integrated =F extracted ·U flags Get integrated medical features and marker data F integrated , where U flags It is an emergency sign. extracted is the extracted medical feature set.

7. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The steps for obtaining the key feature matching results are specifically as follows: Based on the medical characteristics and emergency signs, and compared with historical medical data, the formula is used: Generate the difference comparison result RY, where F integrated [i] is the i-th integrated medical feature and marker data, H data [i] is the corresponding i-th eigenvalue in the historical data, k Y is the number of features; Based on the difference comparison results, the matching degree is evaluated using the formula: MY=exp(-α Y RY) Get the matching degree MY, where RY is the difference comparison result, α Y It is a tuning parameter used to control the matching sensitivity.

8. The medical image instant feedback system based on edge computing according to claim 1 is characterized in that: The steps for obtaining the feedback parameters for tuning are specifically as follows: Based on the key feature matching results, the error vector is calculated according to the current measurement value and feedback results, using the formula: Get the accuracy evaluation result SD, where FH current [i] is the i-th eigenvalue of the current feedback, OF actual [i] is the i-th eigenvalue of the current observation, UO i is the weight of the i-th feature, pJ is the norm order, n V is the total number of features; Based on the accuracy evaluation results, parameter adjustment is performed, and the hyperbolic tangent function is used to adjust the parameter update degree, using the formula: ZP new =ZP old ·(1+β Z ·tanh(γ Z ·SD-δ Z )) Get the new parameter setting ZP new , among which ZP old is the original parameter setting, β Z , γ Z and δ Z is the coefficient for adjusting sensitivity, scale, and bias, and SD is the accuracy assessment result; Apply the new parameter settings, optimize the data processing flow and priority queue, and update the configuration based on the feedback to obtain the tuned feedback parameters.

Images