A method for identifying lesions in magnetic resonance images based on image processing
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LINGYING (SUZHOU) MEDICAL TECH CO LTD
- Filing Date
- 2026-03-24
- Publication Date
- 2026-06-23
AI Technical Summary
Existing MRI image lesion identification systems have low accuracy and inconsistent identification results for different types and locations of lesions, resulting in low identification efficiency and easy introduction of human error.
A correlation library of human body detection model and lesion recognition model was established. Through simulation verification and comprehensive model optimization, combined with the accurate lesion recognition correlation model, the accurate recognition of lesions in MRI images was achieved.
It improves the accuracy and detection rate of lesion identification in MRI images, reduces human error, and ensures the accuracy and efficiency of lesion identification by combining comprehensive and precise models.
Smart Images
Figure CN122265211A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of magnetic resonance imaging lesion recognition technology, specifically a magnetic resonance imaging lesion recognition method based on image processing. Background Technology
[0002] Magnetic resonance imaging (MRI) is a non-invasive medical imaging technique that provides high soft tissue contrast images, which are of great value for disease diagnosis and differential diagnosis. However, the identification and annotation of lesions in MRI images still requires manual operation, which is time-consuming, labor-intensive, and prone to human error. Therefore, developing a system that can automatically or semi-automatically identify and annotate lesions in MRI images has significant clinical value.
[0003] In existing technologies, several automated lesion identification systems have been developed and applied. These systems are mainly based on deep learning and artificial intelligence technologies, achieving automatic lesion identification and annotation through training and learning on a large number of MRI images. However, these systems still have some problems, such as low identification accuracy and inconsistent identification results for lesions of different types and locations. Therefore, in order to improve the accuracy of lesion identification, this invention provides a magnetic resonance image lesion identification method based on image processing. Summary of the Invention
[0004] To address the problems of the aforementioned solutions, this invention provides a magnetic resonance imaging lesion identification method based on image processing, which can more accurately and quickly identify and label lesion regions in MRI images, thereby improving the lesion detection rate and labeling efficiency.
[0005] The objective of this invention can be achieved through the following technical solutions: A method for lesion identification in magnetic resonance imaging based on image processing, the method comprising: Step 1: Establish a human body detection model, and establish lesion recognition models based on the human body detection model; filter and associate the lesion recognition models to form a model library corresponding to the human body detection model; Furthermore, methods for establishing human detection models include: Obtain the magnetic resonance imaging (MRI) detection range, and determine the various body parts and corresponding examination items to be examined on the patient based on the MRI detection range; A human body model is established, and the corresponding parts of each human body examination site are marked in the human body model. The corresponding examination items are also marked at the corresponding human body examination sites in the human body model. The current human body model is marked as a human body detection model.
[0006] Furthermore, methods for screening and associating various lesion identification models include: Acquire historical lesion identification data, and set corresponding simulation data based on the historical lesion identification data. The simulation data includes the corresponding magnetic resonance images, human examination sites and examination items, and lesion results. The simulated data are classified to form several simulation sets; the lesion identification model is simulated and verified using each simulation set to obtain a corresponding result set, which consists of each simulated value. Based on the result set, determine the corresponding simulated representative value; identify each simulated representative value corresponding to each human body examination site and examination item, and mark the lesion identification model with the highest simulated representative value as the accurate lesion identification association model for the corresponding human body examination site and examination item. By analogy, precise lesion identification association models for each human body examination site and examination item are obtained; and each of the obtained precise lesion identification association models is associated with the corresponding markers in the human body detection model. Establish a simulation library and input each related simulation into the model library for storage.
[0007] Furthermore, the methods for simulating and validating the lesion identification model using various simulation sets include: The lesion identification model is simulated sequentially using each of the simulation data in the simulation set to obtain the simulation results of each of the simulation data in the simulation set. The simulation results are evaluated based on the lesion results corresponding to the simulation data and the preset indicator items to obtain the corresponding simulation values; The simulated values corresponding to the simulation set are integrated to form the result set obtained by the lesion identification model after simulation by the simulation set.
[0008] Furthermore, methods for evaluating simulation results based on the lesion findings corresponding to the simulation data and the preset indicator items include: Calculate the similarity between the simulation results and the corresponding lesion results in the simulation data on each indicator item; According to the formula Calculate the simulated values corresponding to each simulation result; In the formula: Dz is the simulated value; λi is the weight coefficient of the corresponding indicator item, i=1,2,...,n, where n is a positive integer; Szi is the similarity of the corresponding indicator item; H(k) is the judgment model.
[0009] Furthermore, the expression for the judgment model is: ; In the formula: k is the comparison result between the simulation result and the lesion result, and U is the set of various inconsistencies in the preset results; It is an empty set.
[0010] Step 2: Determine the comprehensive model and optimize the equipment based on the model library; Further, the method for determining the comprehensive model includes: Identify the simulation representative values of each lesion recognition model for each human body examination part and examination item, and calculate the corresponding comprehensive value according to the formula ZH = ∑Dz j Calculate the corresponding comprehensive value; In the formula: ZH is the comprehensive value; j = 1, 2,..., m, m is a positive integer, j represents the corresponding human body examination part and examination item; Dz j Is the simulation representative value of the corresponding human body examination part and examination item; Select the lesion recognition model with the highest comprehensive value as the comprehensive model.
[0011] Further, the method for optimizing the equipment based on the model library includes: Establish a configuration plan library, which is used to store various configuration plans; Obtain equipment data, input the equipment data into the configuration plan library for matching, obtain each candidate plan, screen the obtained candidate plans, obtain the target configuration plan, and optimize the equipment according to the target configuration plan.
[0012] Further, the method for screening the obtained candidate plans includes: Analyze the probability of artifact occurrence and misdiagnosis rate corresponding to the application of each candidate plan; Estimate the implementation cost corresponding to each candidate plan; According to the formula PTY = b1×(e -CB +1)+b2×[1÷(YL×WL)] 1 / 2 Calculate the priority value of the corresponding candidate plan; In the formula: PTY is the priority value; b1 and b2 are both proportionality coefficients, and the value range is 0 < b1 < 1, 0 < b2 < 1; CB is the implementation cost; YL is the artifact occurrence rate; WL is the misdiagnosis rate; Select the candidate plan with the highest priority value as the target configuration plan.
[0013] Step 3: Obtain the magnetic resonance image that needs to be subjected to lesion recognition, and identify the patient information corresponding to the magnetic resonance image; match the corresponding precise lesion recognition association model according to the patient information and the human body detection model; Step 4: Perform lesion recognition on the magnetic resonance image through the comprehensive model, obtain the corresponding comprehensive recognition result, and process the magnetic resonance image based on the obtained comprehensive recognition result; mark the processed magnetic resonance image as the target image; Step 5: Perform lesion recognition on the target image by the precise lesion recognition association model to obtain the corresponding lesion recognition result.
[0014] Furthermore, it also includes step six: associating and verifying the obtained comprehensive identification result with the lesion identification result to obtain the corresponding verification value. When the verification value is lower than the threshold X1, an abnormal alarm is triggered.
[0015] Compared with the prior art, the beneficial effects of the present invention are: By setting up precise lesion identification association models and a comprehensive model, more accurate identification of lesions in magnetic resonance images can be achieved. The comprehensive model is the lesion identification model with a fixed configuration. The comprehensive model is used for preliminary lesion identification. After the lesion is identified, the corresponding precise lesion identification association model is used for further identification to obtain a more accurate precise lesion identification association model. This allows for preliminary identification by the general comprehensive model, which marks corresponding artifacts, lesions, etc., and removes corresponding interference factors. Then, the precise lesion identification association model is used for accurate identification, which improves the accuracy of lesion identification and eliminates interference factors as much as possible. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0019] like Figure 1 As shown, a method for lesion identification in magnetic resonance imaging based on image processing is described, the method comprising: Step 1: Obtain the corresponding human examination sites and examination items for each lesion, such as cerebral infarction, abdominal tumor, gallstone detection, etc.; establish a human body model, that is, a stereoscopic display data model established using existing 3D technology; mark the human body model according to the obtained human examination sites and corresponding examination items to form a corresponding human detection model. Based on the human body detection model, corresponding lesion recognition models are established. That is, according to the examination sites and examination items of each human body, existing lesion recognition models that can identify lesions in their magnetic resonance images are obtained, as well as new lesion recognition models established manually, resulting in several lesion recognition models. Each lesion recognition model is marked with the corresponding magnetic resonance images of the human body examination sites and examination items that it can recognize. Various existing lesion recognition models are fully utilized, and several targeted lesion recognition models can also be established manually. That is, lesion recognition models for specific detection and recognition can be established. The obtained lesion recognition models are filtered and associated to form a model library corresponding to the human body detection model.
[0020] For example, a lesion identification model can be established manually in the following manner, or other methods can be used to establish the corresponding lesion identification model.
[0021] A sample dataset was created, consisting of source images and manually labeled images. The source images were the original MRI images, while the manually labeled images were format-converted and manually annotated with lesions. The ratio of source images to manually labeled images in the sample dataset was 2:1. The images in the manually labeled images were binarized and saved in single-channel format. The image sample set and the manually labeled images were then split proportionally to form a second image sample set and a second manually labeled image set. A Linknet network model was built based on the Linknet network structure within the PyTorch deep learning framework. The parameters of the Linknet network model were set, and the second image sample set and the second manually labeled image set were input into the Linknet network model. The Linknet network model was trained using the PyTorch deep learning framework. During training, multiple models were saved, and the model with the smallest error was selected as the lesion recognition model using the validation set data.
[0022] Methods for screening and associating various lesion identification models include: Using a large amount of historical lesion identification data, corresponding simulation data is set up. The simulation data includes corresponding magnetic resonance images, human examination sites and examination items, and lesion results. That is, several sets of simulation data are formed using existing historical lesion identification data, where the lesion results are the accurate lesion identification results of the magnetic resonance images in the simulation data. The simulation data is categorized according to the corresponding human examination sites and examination items to form a simulation set for each human examination site and examination item; the lesion identification model is simulated and verified through each simulation set to obtain the corresponding result set, which consists of the simulation values corresponding to each simulation data. Based on the obtained result set, determine the simulated representative value for the corresponding lesion identification model. The simulated representative value is the simulated representative value in the result set. The corresponding simulated representative value can be determined by calculating the mean, standard deviation, mode, etc. The simulation representative values corresponding to each human body examination site and examination item are identified. The lesion identification model with the highest simulation representative value is marked as the accurate lesion identification association model for that human body examination site and examination item. That is, the accurate lesion identification association model is the most accurate in identifying lesions in the corresponding magnetic resonance images of that human body examination site and examination item. This process is repeated to obtain the accurate lesion identification association models for each human body examination site and examination item. The obtained accurate lesion identification association models are then associated with the corresponding human body examination site and examination item markings in the human body detection model. Establish a simulation library and input each related simulation into the model library for storage.
[0023] Methods for simulating and validating lesion identification models using various simulation sets include: The lesion identification model is simulated sequentially using the corresponding simulation data in each simulation set to obtain the simulation results of each simulation data in the simulation set. The simulation results are the identification status of the lesion, such as whether the lesion is identified, the location, shape, volume and other indicators of the lesion. Preset various indicators, which are used to judge whether the lesion identification is accurate. They can also be set according to the purpose of detailed lesion data, such as volume, signal intensity, morphological features, texture features, etc. The specifics can be adjusted according to the actual situation. Determine the impact of each indicator on the diagnosis of the corresponding disease when the identification is wrong through historical data, such as statistical error rate, number proportion, etc., and determine the weight coefficient of each indicator based on the magnitude of the impact of each indicator. Calculate the similarity of the simulation results with the corresponding lesion results in the simulation data on each indicator item, according to the formula. Calculate the simulated values corresponding to each simulation result; In the formula: Dz is the simulated value; λi is the weight coefficient of the corresponding indicator item, i=1,2,...,n, where n is a positive integer; Szi is the similarity of the corresponding indicator item; H(k) is the judgment model, which is established based on the judgment function, that is, it is set according to whether the simulation result and the lesion result are consistent. For example, if the lesion result shows a lesion at the corresponding location, but the simulation result shows no lesion within the corresponding range, the result difference is too large and is considered inconsistent; the expression is: In the formula: k is the comparison result between the simulation result and the lesion result, and U is the set of various inconsistencies in the preset results; It is an empty set.
[0024] The simulated values obtained by simulating the lesion identification model through the simulation set are integrated into the result value.
[0025] Step 2: Determine the comprehensive model and optimize the equipment based on the model library; Determine a comprehensive model, which is the selected lesion identification model. This model can be the lesion identification model built into the MRI equipment, or a lesion identification model selected by other methods, or it can be determined in the following manner. Identify the simulated representative values of each lesion identification model for each human examination site and examination item, based on the formula ZH=∑Dz j Calculate the corresponding comprehensive value; where: ZH is the comprehensive value; j = 1, 2, ..., m, where m is a positive integer, and j represents the corresponding human body examination site and examination item; Dz j These are simulated representative values for corresponding human examination sites and examination items; The lesion identification model with the highest comprehensive value is selected as the comprehensive model. The comprehensive model is used for direct application.
[0026] Equipment optimization based on the model library involves performing optimization analysis based on the identified equipment conditions to configure the model library. This is mainly done by utilizing existing relevant configuration technologies and establishing a corresponding configuration scheme library based on these technologies. The configuration scheme library stores various configuration schemes. The configuration scheme includes relevant data such as configuration methods, applicable scope, and cost estimates; various exemplary configuration methods are as follows: The model is deployed on multiple devices, utilizing their computing resources and storage space to collaboratively complete computational tasks; specialized hardware devices, such as GPUs and FPGAs, are used to accelerate the model's computation process and improve its operating efficiency; caching techniques are used to reduce redundant computations and data access, further improving the model's operating efficiency; and the system is optimized to enhance the devices' parallel processing capabilities and I / O performance, thereby improving the model's operating efficiency and stability. Deploying models on cloud computing services such as Amazon Web Services (AWS), Google Cloud Platform (GCP), and Microsoft Azure provides powerful computing and storage resources that can be dynamically scaled up or down as needed. Alternatively, packaging models into container images and running them in the cloud using container services allows for rapid deployment and scaling, while ensuring version consistency across different environments. Finally, storing the data that the model needs to process in cloud database services such as AWS DynamoDB, Google Cloud Datastore, and Azure CosmosDB can also be effective.
[0027] Obtain device data, input the obtained device data into the configuration scheme library for matching, obtain various configuration schemes that meet the requirements, that is, configuration schemes that satisfy the configuration conditions, and match according to the corresponding applicable scope; Mark the matched configuration schemes as candidate schemes, screen the obtained candidate schemes, obtain the target configuration scheme, and optimize the device according to the obtained target configuration scheme.
[0028] The methods for screening the obtained candidate schemes include: Estimate the implementation costs corresponding to each candidate scheme; Analyze the influence value on magnetic resonance image generation when applying the candidate scheme. The influence value refers to when applying the candidate scheme, according to the implementation situation of the candidate scheme, analyze the influence on magnetic resonance image generation, mainly determine the corresponding influence degree based on artifacts, such as the influence of generating artifacts after supplementing corresponding devices; Combine the corresponding probability of generating artifacts and the influence on doctor diagnosis when such artifacts occur, which can be determined according to the misdiagnosis rate; Then the influence value = [1÷(probability of artifact occurrence × misdiagnosis rate)] 1 / 2 ; Among them, the probability of artifact occurrence and the misdiagnosis rate can both be statistically analyzed using existing historical data or through actual simulation to determine the corresponding probability of artifact occurrence and misdiagnosis rate.
[0029] According to the formula PTY = b1×(e -CB +1)+b2×YR to calculate the priority value of the corresponding candidate scheme; In the formula, PTY is the priority value; b1 and b2 are both proportionality coefficients, and the value range is 0 < b1 < 1, 0 < b2 < 1; CB is the implementation cost; YR is the influence value; Select the candidate scheme with the highest priority value as the target configuration scheme.
[0030] By setting each precise lesion recognition association model and comprehensive model, it is used to achieve more precise recognition of lesions in magnetic resonance images. The comprehensive model is a fixed-configuration lesion recognition model. Through the comprehensive model for preliminary lesion recognition, when a lesion is recognized, the corresponding precise lesion recognition association model associated with the recognized lesion is used for re-recognition to obtain a more precise precise lesion recognition association model; It is convenient to first perform preliminary recognition by the general comprehensive model, mark corresponding artifacts, lesions, etc., remove corresponding interference factors, and then perform precise recognition by the precise lesion recognition association model to improve the recognition accuracy of lesions and尽可能 eliminate interference factors as much as possible.
[0031] Step three: Obtain the magnetic resonance image that needs to be subjected to lesion recognition, as well as the patient information corresponding to the magnetic resonance image; Match the corresponding precise lesion recognition association model in the human body detection model according to the obtained patient information; First, locate the corresponding human body examination site, and then determine the corresponding precise lesion identification association model based on the corresponding examination items.
[0032] Step 4: Use a comprehensive model to identify lesions in the magnetic resonance images and obtain the corresponding comprehensive identification results. Based on the obtained comprehensive identification results, process the magnetic resonance images; such as marking the identified artifacts, lesions, etc.; or eliminating certain interfering factors. The specific processing is carried out according to the preset processing scheme; mark the processed magnetic resonance images as target images. Step 5: The target image is used to identify lesions using the corresponding accurate lesion identification and association model to obtain the corresponding lesion identification results.
[0033] In one embodiment, since the magnetic resonance images have been analyzed by both the comprehensive model and the precise lesion identification association model, resulting in two lesion identification results, these two results can be cross-checked to determine if the difference between them is too large. If the difference is too large, it indicates an anomaly, because even with a difference in accuracy, the difference in identification would not be significant. Therefore, this embodiment also includes step six: cross-checking the obtained comprehensive identification result with the lesion identification result to obtain a corresponding check value. When the check value is lower than the threshold X1, an anomaly alarm is triggered, and the corresponding staff is alerted; otherwise, no corresponding operation is performed.
[0034] The verification value is calculated using the formula for the simulated value, i.e. Mark the corresponding simulated value as the verification value.
[0035] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.
[0036] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for identifying lesions in magnetic resonance imaging based on image processing, characterized in that, The methods include: Step 1: Establish a human body detection model, and establish a lesion identification model based on the human body detection model; The lesion identification models are screened and associated to form a model library corresponding to the human body detection model; Step 2: Determine the comprehensive model for lesion identification and optimize the equipment based on the model library; Step 3: Acquire the magnetic resonance images for lesion identification, and identify the patient information corresponding to the magnetic resonance images; match the corresponding accurate lesion identification association model based on the patient information and the human body detection model; Step 4: Identify lesions in the magnetic resonance image using the comprehensive model to obtain the corresponding comprehensive identification result, and process the magnetic resonance image based on the obtained comprehensive identification result; The processed magnetic resonance image is labeled as the target image; Step 5: The precise lesion identification and association model is used to identify lesions in the target image to obtain the corresponding lesion identification results.
2. The method for identifying lesions in magnetic resonance images based on image processing according to claim 1, characterized in that, Methods for establishing human body detection models include: Obtain the magnetic resonance imaging (MRI) detection range, and determine the various body parts and corresponding examination items to be examined on the patient based on the MRI detection range; A human body model is established, and the corresponding parts of each human body examination site are marked in the human body model. The corresponding examination items are also marked at the corresponding human body examination sites in the human body model. The current human body model is marked as a human body detection model.
3. The method for identifying lesions in magnetic resonance images based on image processing according to claim 2, characterized in that, Methods for screening and associating various lesion identification models include: Acquire historical lesion identification data, and set corresponding simulation data based on the historical lesion identification data. The simulation data includes the corresponding magnetic resonance images, human examination sites and examination items, and lesion results. The simulated data are classified to form several simulation sets; the lesion identification model is simulated and verified using each simulation set to obtain a corresponding result set, which consists of each simulated value. Based on the result set, determine the corresponding simulated representative value; identify each simulated representative value corresponding to each human body examination site and examination item, and mark the lesion identification model with the highest simulated representative value as the accurate lesion identification association model for the corresponding human body examination site and examination item. By analogy, precise lesion identification association models for each human body examination site and examination item are obtained; and each of the obtained precise lesion identification association models is associated with the corresponding markers in the human body detection model. Establish a simulation library and input each related simulation into the model library for storage.
4. The method for identifying lesions in magnetic resonance images based on image processing according to claim 3, characterized in that, Methods for simulating and validating lesion identification models using various simulation sets include: The lesion identification model is simulated sequentially using each of the simulation data in the simulation set to obtain the simulation results of each of the simulation data in the simulation set. The simulation results are evaluated based on the lesion results corresponding to the simulation data and the preset indicator items to obtain the corresponding simulation values; The simulated values corresponding to the simulation set are integrated to form the result set obtained by the lesion identification model after simulation by the simulation set.
5. The method for identifying lesions in magnetic resonance images based on image processing according to claim 4, characterized in that, Methods for evaluating simulation results based on the lesion findings corresponding to the simulation data and the preset indicators include: Calculate the similarity between the simulation results and the corresponding lesion results in the simulation data for each index item; According to the formula Calculate the simulated values corresponding to each simulation result; In the formula: Dz is the simulation value; λi is the weight coefficient of the corresponding index item, i = 1, 2,..., n, n is a positive integer; Szi is the similarity of the corresponding index item; H(k) is the judgment model.
6. The method for identifying lesions in magnetic resonance images based on image processing according to claim 5, characterized in that, The expression for the judgment model is: ; In the formula: k is the comparison result between the simulation result and the lesion result, and U is the set of various inconsistencies in the preset results; It is an empty set.
7. The method for identifying lesions in magnetic resonance images based on image processing according to claim 5, characterized in that, The method for determining the comprehensive model includes: Identify the simulated representative values of each lesion identification model for each human examination site and examination item, according to the formula ZH=∑Dz j Calculate the corresponding composite value; In the formula: ZH is the comprehensive value; j = 1, 2, ..., m, where m is a positive integer, and j represents the corresponding human body examination site and examination item; Dz j These are simulated representative values for corresponding human examination sites and examination items; Select the lesion recognition model with the highest comprehensive value as the comprehensive model.
8. The method for identifying lesions in magnetic resonance images based on image processing according to claim 1, characterized in that, The method for optimizing the device based on the model library includes: Establish a configuration plan library, which is used to store various configuration plans; Obtain device data, input the device data into the configuration plan library for matching, obtain each candidate plan, screen the obtained candidate plans, obtain the target configuration plan, and optimize the device according to the target configuration plan.
9. The method for identifying lesions in magnetic resonance images based on image processing according to claim 8, characterized in that, The method for screening the obtained candidate plans includes: Analyze the probability of artifact occurrence and misdiagnosis rate corresponding to the application of each candidate plan; Estimate the implementation cost corresponding to each candidate plan; According to the formula PTY=b1×(e -CB +1)+b2×[1÷(YL×WL)] 1 / 2 Calculate the priority value of the corresponding candidate options; In the formula: PTY is the priority value; b1 and b2 are both proportionality coefficients, and the value range is 0 < b1 < 1, 0 < b2 < 1; CB is the implementation cost; YL is the artifact occurrence rate; WL is the misdiagnosis rate; Select the candidate plan with the highest priority value as the target configuration plan.
10. The method for identifying lesions in magnetic resonance images based on image processing according to claim 1, characterized in that, It further includes Step Six: Correlate and verify the obtained comprehensive recognition result with the lesion recognition result, obtain the corresponding verification value, and when the verification value is lower than the threshold X1, an abnormal alarm is issued.