CN103984861B - Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device - Google Patents
Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device Download PDFInfo
- Publication number
- CN103984861B CN103984861B CN201410202258.XA CN201410202258A CN103984861B CN 103984861 B CN103984861 B CN 103984861B CN 201410202258 A CN201410202258 A CN 201410202258A CN 103984861 B CN103984861 B CN 103984861B
- Authority
- CN
- China
- Prior art keywords
- hemodialysis
- pressure
- complication
- fuzzy
- dialysis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- External Artificial Organs (AREA)
Abstract
The invention relates to a fuzzy neural network and rule-based expert system fused online hemodialysis monitoring method. The method comprises the steps: realizing the construction of a hemodialysis machine complication diagnosis system in a manner that fuzzy logics are utilized, a neural network is taken as a core and an expert system is combined with actual situations of a hemodialysis process; establishing a fuzzy membership degree function for patient blood pressure x1, heart rate x2, hemodialysis machine inlet arterial pressure x3, hemodialysis machine inlet venous pressure x4, blood color sampling x5, hemodialysis machine transmembrane pressure x6, body temperature x7 and blood sugar x8, a corresponding relationship between process variables and hemodialysis complications when hemodialysis patient excessive-dehydration y1, hemodialysis pipeline blood coagulation y2, thrombus y3, hemolysis y4, heat sourced reaction y5, glycopenia y6 and imbalance syndrome y7 arise, and a fuzzy rule base. A trained network can be used for quickly and accurately diagnosing the possibility of the complications and provides technical feasibility for the remote access treatment of doctors and the reduction of manual intervention.
Description
Technical field
The present invention relates to a kind of on-line monitoring method of hemodialysis is and in particular to a kind of merge fuzzy neural network and rule
The then hemodialysis on-line monitoring method of type specialist system.
Background technology
Hemodialysis is renal failure safely and effectively one of alternative medicine, is widely used to clinic.Because hemodialysis is one
Item long therapeutic procedure, so require higher to security reliability.However, leaking blood, blood coagulation in current hemodialysis, losing blood and blood
Flow the complication such as not good enough to happen occasionally, if it is determined that then can directly jeopardize the personal safety of patient not in time.
At present both at home and abroad in terms of to hemodialysis monitoring, mainly still adopt manual type, this requires that medical personnel should
There are the operating technology of consummation and abundant clinical position experience, possess certain technical management ability again, could maintaining treatment
Process safety is effective, but manual monitoring has certain randomness, limitation and subjectivity.Especially in high flux hemodialysis mould
Under formula, due to the raising of velocity of blood flow, system mode at any time it may happen that change, medical personnel in busy work, unavoidably
Mistake estimation to state occurring, thus affecting the promptness of emergency processing, being that patient brings unnecessary misery.With calculating
Machine and automatic control technology also achieve online prison to a certain extent in the application of Modern Medical Field and development, hemo system
Survey, but its application is not only extremely limited, and its simple condition adjudgement and single emergency processing are likely to occur
Clinical problem.For example: extracorporeal circulation at present is in integrated state with draining blood vessel interior circulation, at draining blood vessel or extracorporeal circulation
When occurring incomplete obstruction to cause blood stream smooth, hemo system according to judging and can do corresponding action, that is, suspend whole
Hemodialysis circulates, but if time out is slightly long, will cause hemodialysis device or perfusion device blocking, and then lead to therapeutic process to be forced to stop
Only.Especially for the bad patient of vascular conditions, this situation will make its therapeutic effect become very poor.
In today of systematic treatment with subjects medical skill development, hemodialysis on-line monitoring technique is also to automatization and intelligence
Change direction to develop.Neutral net has acquisition knowledge, the ability of parallel processing, stored knowledge and self study, but it does not simply fail to
Explain the reasoning process of oneself, and because hemodialysis are more complicated, correlated variabless are numerous and INFORMATION OF INCOMPLETE degree is higher, from
And limit its application in hemodialysis on-line monitoring.Specialist system contains knowledge and the warp of certain domain expert's level substantial amounts of
Test, this field question can be processed using the method for the knowledge of human expert and solve problem.That is, specialist system simulation
The decision making process of human expert, so as to solve those need human expert process challenge, therefore can be used for hemodialysis online
Monitoring, but the work efficiency of its search strategy is relatively low, and there is substantial amounts of exhaustive search.Additionally, specialist system do not possess from
The ability of experience learning, it is impossible to automatically change knowledge base, also cannot be realized well in hemodialysis on-line monitoring and use.
Content of the invention
In order to solve the above problems, it is an object of the invention to provide a kind of merge fuzzy neural network and regular pattern composite expert
The hemodialysis on-line monitoring method of system, completes the foundation in initial knowledge storehouse using the correlation technique of fuzzy logic and specialist system,
In conjunction with hemodialysis practical situation, by the analysis to correlated process signal, set up corresponding complication knowing in the hemodialysis stage
Know storehouse, the neutral net being built using knowledge base training, thus automated decision system status export relevant information, also may be used
In time according to the situation more new knowledge base of patient, realize the purpose of in emergency circumstances accurately anticipation.
In order to achieve the above object, the technical solution adopted in the present invention is:
A kind of hemodialysis on-line monitoring method merging fuzzy neural network and regular pattern composite specialist system, using fuzzy logic,
With neutral net as core, combine the practical situation of hemodialysis using specialist system, anticipation is carried out to the complication occurring, step
Suddenly as follows:
Step 1: first to process variable: blood pressure, heart rate, the sound pulse pressure of hemo system input/output port, transmembrane pressure, body
Temperature, blood glucose, color sampling carry out Fuzzy Processing, according to expertise, process variable are divided into different fuzzy sets and merge tax
Corresponding degree of membership is given to make a distinction, the fuzzy membership establishment method of detailed process variable is as follows:
1. patients' blood's value x in hemodialysis1: { pressure value x1It is set to 1 less than 90mmhg；Pressure value x1Higher than 160mmhg
It is set to 0.7；Pressure value x1It is set to 0.3 in 140-160mmhg；Pressure value x1It is set to 0 in 90-140mmhg }
2. Heart Rate value x in hemodialysis2: { heart rate x2It is set to 1 higher than 130 beats/min；Heart rate x2At 100-130 beat/min
It is set to 0.8；Heart rate x2It is set to 0 at 60-100 beat/min；Heart rate x2It is set to 0.4 below 60 beats/min }
3. in hemodialysis hemo system entrance arterial pressure measurement value x3: { arterial pressure x3It is set to 0.5 in 0.5-3.0psi；Dynamic
Pulse pressure x3It is set to 0.9 in 0-0.5psi；Arterial pressure x3It is set to 1 less than 0psi；Arterial pressure x3It is set to 0 in 3.0-5.0psi }
4. in hemodialysis hemo system entrance venous pressure measured value x4: { venous pressure x4It is set to 0.8 in 2.5-3.5psi；Quiet
Pulse pressure x4It is set to 0.5 in 1.0-2.5psi；Venous pressure x4It is set to 1 more than 3.5psi；Venous pressure x4It is set to 0 in -0.5-1.0psi }
5. color sampling x in hemodialysis5: { color sampling takes on a red color and is set to 0；Vinicolor is set to 0.9；Light red jaundice
It is set to 1 }
6. hemodialysis device transmembrane pressure measured value x6: { transmembrane pressure x6It is set to 0.7 less than 0psi；Transmembrane pressure x6Between 0-0.3psi
It is set to 0.4；Transmembrane pressure x6It is set to 0 between 0.1-1psi；Transmembrane pressure x6It is set to 1 in other scopes }
7. core body temperature measurements x in hemodialysis7: { body temperature x7It is set to 0 below 37.5 DEG C；Body temperature x7Fixed at 37.5-38 DEG C
For 0.6；Body temperature x7It is set to 1 more than 38 DEG C }
8. patient blood glucose's content measurement value x in hemodialysis8: { patient blood glucose's content x8It is set to 0 in 3.9-6.1mmol/l；
Blood sugar content x8It is set to 0.3 in 2.8-3.9mmol/l；Blood sugar content x8It is set to 0.8 less than 2.8mmol/l；Blood sugar content x8High
It is set to 1 in 6.1mmol/l }
Step 2: the complication being likely to occur in hemodialysis is: y1: hemodialysis patients dehydration is excessive；y2: hemodialysis pipeline coagulates
Blood；y3: thrombosis；y4: haemolysis；y5: heat source response；y6: hypoglycemia；y7: imbalance syndrome.According to expertise knowledge, set up step
Process variable in rapid 1 and hemodialysis patients dehydration excess, hemodialysis pipeline blood coagulation, thrombosis, haemolysis, heat source response, hypoglycemia and mistake
The complication such as weighing apparatus syndrome are in the expert knowledge library in hemodialysis stage.Obtain between the process variable shown in table 1 and Complication of Hemodialysis
Corresponding relation and fuzzy rule base.Wherein, process variable is regular former piece, and complication is consequent.
Corresponding relation between process variable and Complication of Hemodialysis in table 1 hemodialysis
Step 3: build neural network model, including to the selection of network various parameters and the foundation of establishment, model and right
Network model is trained.The present invention is using the multilayer feedforward neural network by Back Propagation Algorithm training, hereinafter referred to as bp
Neutral net, the bp neutral net of use has Three Tiered Network Architecture, using the regular former piece in step 2 as bp neutral net
Input layer, input layer has 8 process variables, corresponding 8 nodes；Consequent in step 2 is as the output of bp neutral net
Layer, output layer has 7 complication, corresponding 7 nodes；The establishment of hidden layer neuron number is using formula (1) and formula (2):
Wherein, nimplicitFor hidden neuron number；ninputFor input layer number；noutputFor output layer nerve
First number, the value of a is 1～10.Herein, determine that node in hidden layer is 14.In described hidden layer, network weight is according to error
Back propagation principle carries out dynamic error correction determination, and the training of neutral net adopts quick self-adapted algorithm, using with attached
Plus the gradient descent method traingdx function of activity level method and adjusting learning rate is as the training function of network, e-learning letter
Number is the gradient descent method function (learngdm) with momentum term, and performance function adopts sum of squared errors function (sse) and sets
It is set to 0, initial learn speed is 0.01, momentum constant is 0.95, and maximum cycle is 2000 times, and each layer transfer function is
The s type function (log-sigmoid) taken the logarithm.
Using the rule (table 1) in expert system knowledge base as the input of network and output, complete the instruction to neutral net
Practice.
Step 4: the model in hemodialysis on-line monitoring, the actual monitoring value input of process variable being trained, check model
Output result.This monitoring method corresponding Complication of Hemodialysis output result is in the range of 0-1, true by the degree of membership between 0 to 1
The probability size that vertical complication occurs, the symbolical meaningses such as table 3 of probability degree of membership in complication.
Table 3: the degree of membership of probability in complication
| Degree of membership | The probability that complication occurs |
| 0.8-0.1 | Occur |
| 0.6-0.8 | It is likely to occur |
| 0.4-0.6 | Uncertain |
| 0.2-0.4 | May not occur |
| 0-0.2 | Do not occur |
The different complication accurately portrayed and same process variable caused to process variable are realized using fuzzy logic
Deep discrimination.
Establish fuzzy membership function using expertise knowledge, realize the process variable value to input bp neutral net
Fuzzy processing, disclosure satisfy that the needs of bp neutral net input data, again pretreatment need not be carried out to input data；This
Outward, the self-learning capability of neutral net makes renewal and the increase of knowledge base, improves the expansion of system, thus realizing to more
How larger range of complication diagnoses.
Using the small-sized ddc system that is made up of plc and computer of hemodialysis system, plc carries out pre- to the data of collection
Process, and transfer to computer analysis output result.
The present invention carries out the establishment of the fuzzy membership function of above-mentioned 8 process variables respectively by expertise, and
The design of hemodialysis variable-Complication of Hemodialysis sample set, builds initial knowledge storehouse；And determine neural network topology structure and
Network parameter, merges fuzzy neural network and regular pattern composite specialist system.In a particular application, the data of neural network expert system
The data that real-time detection arrives is sent into specialist system by processing module, and specialist system calls data and real-time detection in knowledge base
Data is compared, by the supposition of ANN Reasoning machine and analysis finally judge system with the presence or absence of Complication of Hemodialysis and
There is a possibility that size；Hemodialysis efficiency is ensured with the effect reaching hemodialysis safety and stability simultaneously.The network energy training
Enough complication is made quickly and accurately judges, is easy to medical personnel and takes measures or as the foundation recovered.Solving blood
When the various problems such as leaking blood, blood coagulation during thoroughly, lose blood, inevitably there is randomness, limitation and subjectivity in artificial experience
Property, by hemodialysis safety be increased to the on-line monitoring that complication occurs.
Brief description
Accompanying drawing is a kind of structural frames of the hemodialysis on-line monitoring method merging fuzzy neural network and regular pattern composite specialist system
Figure.
Specific embodiment
With reference to example, the present invention will be described in more detail.
As shown in drawings, a kind of hemodialysis on-line monitoring side merging fuzzy neural network and regular pattern composite specialist system of the present invention
Method.Complete the foundation in initial knowledge storehouse using the correlation technique of fuzzy logic and specialist system, built using knowledge base training
Neutral net, can automated decision system status export corresponding information, also can in time according to patient situation voluntarily or
More new knowledge base under manual intervention, makes up to the purpose of accurate anticipation in emergency circumstances.Step is as follows:
Step 1: first to process variable: blood pressure, heart rate, the sound pulse pressure of hemo system input/output port, transmembrane pressure, body
Temperature, blood glucose, color sampling carry out Fuzzy Processing, according to expertise, process variable are divided into different fuzzy sets and merge imparting
Corresponding degree of membership makes a distinction, and the fuzzy membership establishment method of detailed process variable is as follows:
1. patients' blood's value x in hemodialysis1: { pressure value x1It is set to 1 less than 90mmhg；Pressure value x1Higher than 160mmhg
It is set to 0.7；Pressure value x1It is set to 0.3 in 140-160mmhg；Pressure value x1It is set to 0 in 90-140mmhg }
2. Heart Rate value x in hemodialysis2: { heart rate x2It is set to 1 higher than 130 beats/min；Heart rate x2At 100-130 beat/min
It is set to 0.8；Heart rate x2It is set to 0 at 60-100 beat/min；Heart rate x2It is set to 0.4 below 60 beats/min }
3. in hemodialysis hemo system entrance arterial pressure measurement value x3: { arterial pressure x3It is set to 0.5 in 0.5-3.0psi；Dynamic
Pulse pressure x3It is set to 0.9 in 0-0.5psi；Arterial pressure x3It is set to 1 less than 0psi；Arterial pressure x3It is set to 0 in 3.0-5.0psi }
4. in hemodialysis hemo system entrance venous pressure measured value x4: { venous pressure x4It is set to 0.8 in 2.5-3.5psi；Quiet
Pulse pressure x4It is set to 0.5 in 1.0-2.5psi；Venous pressure x4It is set to 1 more than 3.5psi；Venous pressure x4It is set to 0 in -0.5-1.0psi }
5. color sampling x in hemodialysis5: { color sampling takes on a red color and is set to 0；Vinicolor is set to 0.9；Light red jaundice
It is set to 1 }
6. hemodialysis device transmembrane pressure measured value x6: { transmembrane pressure x6It is set to 0.7 less than 0psi；Transmembrane pressure x6Between 0-0.3psi
It is set to 0.4；Transmembrane pressure x6It is set to 0 between 0.1-1psi；Transmembrane pressure x6It is set to 1 in other scopes }
7. core body temperature measurements x in hemodialysis7: { body temperature x7It is set to 0 below 37.5 DEG C；Body temperature x7Fixed at 37.5-38 DEG C
For 0.6；Body temperature x7It is set to 1 more than 38 DEG C }
8. patient blood glucose's content measurement value x in hemodialysis8: { patient blood glucose's content x8It is set to 0 in 3.9-6.1mmol/l；
Blood sugar content x8It is set to 0.3 in 2.8-3.9mmol/l；Blood sugar content x8It is set to 0.8 less than 2.8mmol/l；Blood sugar content x8It is higher than
6.1mmol/l is set to 1 }
Step 2: the complication being likely to occur in hemodialysis is: y1: hemodialysis patients dehydration is excessive；y2: hemodialysis pipeline coagulates
Blood；y3: thrombosis；y4: haemolysis；y5: heat source response；y6: hypoglycemia；y7: imbalance syndrome.According to expertise knowledge, set up step
Process variable in rapid 1 and hemodialysis patients dehydration excess, hemodialysis pipeline blood coagulation, thrombosis, haemolysis, heat source response, hypoglycemia, unbalance
The complication such as syndrome are in the expert knowledge library in hemodialysis stage.Obtain between the process variable shown in table 1 and Complication of Hemodialysis
Corresponding relation and fuzzy rule base.Wherein, process variable is regular former piece, and complication is consequent.
Corresponding relation between process variable and Complication of Hemodialysis in table 1 hemodialysis
Step 3: build neural network model, including to the selection of network various parameters and the foundation of establishment, model and right
Network model is trained.The present invention is using multilayer feedforward neural network (the abbreviation bp nerve by Back Propagation Algorithm training
Network), the bp neutral net of use has Three Tiered Network Architecture, using the regular former piece in step 2 as neutral net input
Layer, input layer has 8 process variables, corresponding 8 nodes；Consequent in step 2, as the output layer of neutral net, exports
Layer has 7 complication, corresponding 7 nodes；The establishment of hidden layer neuron number is using formula (1) and formula (2):
Wherein, nimplicitFor hidden neuron number；ninputFor input layer number；noutputFor output layer nerve
First number, the value of a is 1～10.Herein, determine that the number of hidden nodes is 14.In described hidden layer, network weight is anti-according to error
Carry out dynamic error correction determination to propagation principle, the training of neutral net adopts quick self-adapted algorithm, using with additional
The gradient descent method traingdx function of activity level method and adjusting learning rate is as the training function of network, e-learning function
It is the gradient descent method function (learngdm) with momentum term, performance function adopts sum of squared errors function (sse) and sets
For 0, initial learn speed is 0.01, and momentum constant is 0.95, and maximum cycle is 2000 times, and each layer transfer function is and takes
The s type function (log-sigmoid) of logarithm.
Using the rule (table 1) in expert system knowledge base as the input of network and output, complete the instruction to neutral net
Practice.
Step 4: the network in actual applications input of process variable sampled value being trained, check network output result.
Table 2 is hospital's hemodialysis collection in worksite data, determines following 8 process variables according to fuzzy membership establishment method in step 1
Degree of membership.
Table 2 actual samples sample value
Hemodialysis on-line monitoring method using above-mentioned fusion fuzzy neural network and regular pattern composite specialist system judges that scene is adopted
The corresponding complication occurring of the process variable of collection and its degree of membership, its output result is shown in Table 4.
Table 4 output result
| Complication | y1 | y2 | y3 | y4 | y5 | y6 | y7 |
| Output is subordinate to angle value | 0.0044 | 0.000 | 0.0066 | 0.9741 | 0.0012 | 0.0000 | 0.1435 |
In this model, the corresponding output result of Complication of Hemodialysis, in the range of 0-1, is established simultaneously by the degree of membership between 0 to 1
Send out the probability size that disease occurs, complication has a symbolical meaningses such as table 3 of the degree of membership of probability:
Table 3: the degree of membership of probability in complication
| Degree of membership | The probability that complication occurs |
| 0.8-0.1 | Occur |
| 0.6-0.8 | It is likely to occur |
| 0.4-0.6 | Uncertain |
| 0.2-0.4 | May not occur |
| 0-0.2 | Do not occur |
Defined by degree of membership in table 3 and know, the complication diagnostic result of this data sampling is:
| Complication | y1 | y2 | y3 | y4 | y5 | y6 | y7 |
| Output is subordinate to angle value | 0.0044 | 0.0000 | 0.0066 | 0.9741 | 0.0012 | 0.0000 | 0.1435 |
| Judged result | Do not occur | Do not occur | Do not occur | Occur | Do not occur | Do not occur | Do not occur |
Therefore, using merging the hemodialysis on-line monitoring method of fuzzy neural network and regular pattern composite specialist system to collection in worksite
The complication diagnostic result of data is: haemolysis in hemodialysis.
Claims (3)
1. merge the hemodialysis on-Line Monitor Device of fuzzy neural network and regular pattern composite specialist system it is characterised in that: this device is
The small-sized ddc system being made up of plc and computer, plc carries out pretreatment to the data of collection, and computer is according to pretreated
Data separate fuzzy logic, with neutral net as core, combines the practical situation of dialysis procedure using specialist system, to occur
Complication carries out anticipation, and this device includes:
Fuzzy membership determining module, its function includes: to process variable: blood pressure, heart rate, the sound of dialysis machine input/output port
Pulse pressure, transmembrane pressure, body temperature, blood glucose, color sampling carry out Fuzzy Processing, are divided into process variable according to expertise different
Fuzzy set merges the corresponding degree of membership of imparting and makes a distinction, and the fuzzy membership establishment method of detailed process variable is as follows:
1. patients' blood's value x in dialysis procedure1: pressure value x1It is set to 1 less than 90mmhg；Pressure value x1It is set to higher than 160mmhg
0.7；Pressure value x1It is set to 0.3 in 140-160mmhg；Pressure value x1It is set to 0 in 90-140mmhg；
2. Heart Rate value x in dialysis procedure2: heart rate x2It is set to 1 higher than 130 beats/min；Heart rate x2It is set at 100-130 beat/min
0.8；Heart rate x2It is set to 0 at 60-100 beat/min；Heart rate x2It is set to 0.4 below 60 beats/min；
3. in dialysis procedure dialysis machine entrance arterial pressure measurement value x3: arterial pressure x3It is set to 0.5 in 0.5-3.0psi；Arterial pressure
x3It is set to 0.9 in 0-0.5psi；Arterial pressure x3It is set to 1 less than 0psi；Arterial pressure x3It is set to 0 in 3.0-5.0psi；
4. in dialysis procedure dialysis machine entrance venous pressure measured value x4: venous pressure x4It is set to 0.8 in 2.5-3.5psi；Venous pressure
x4It is set to 0.5 in 1.0-2.5psi；Venous pressure x4It is set to 1 more than 3.5psi；Venous pressure x4It is set to 0 in -0.5-1.0psi；
5. color sampling x in dialysis procedure5: color sampling takes on a red color and is set to 0；Vinicolor is set to 0.9；Light red jaundice is set to 1；
6. dialyser transmembrane pressure measured value x6: transmembrane pressure x6It is set to 0.7 less than 0psi；Transmembrane pressure x6It is set between 0-0.3psi
0.4；Transmembrane pressure x6It is set to 0 between 0.3-1psi；Transmembrane pressure x6It is set to 1 in other scopes；
7. core body temperature measurements x in dialysis procedure7: body temperature x7It is set to 0 below 37.5 DEG C；Body temperature x7It is set to 0.6 at 37.5-38 DEG C；
Body temperature x7It is set to 1 more than 38 DEG C；
8. patient blood glucose's content measurement value x in dialysis procedure8: patient blood glucose's content x8It is set to 0 in 3.9-6.1mmol/l；Blood glucose contains
Amount x8It is set to 0.3 in 2.8-3.9mmol/l；Blood sugar content x8It is set to 0.8 less than 2.8mmol/l；Blood sugar content x8It is higher than
6.1mmol/l is set to 1；
Process variable and complication corresponding relation determining module, its function includes: the complication being likely to occur in dialysis procedure is:
y1: dialysis patient dehydration is excessive, y2: dialysis pipeline blood coagulation, y3: thrombosis, y4: haemolysis, y5: heat source response, y6: hypoglycemia, y7: lose
Weighing apparatus syndrome；According to expertise knowledge, set up said process variable and be dehydrated excessive, dialyse pipeline blood coagulation, blood with dialysis patient
Bolt, haemolysis, heat source response, hypoglycemia and imbalance syndrome complication, in the expert knowledge library of dialysis stage, obtain shown in table 1
Corresponding relation between process variable and Complication of dialysis and fuzzy rule base；Wherein, process variable is regular former piece, complication
For consequent；
Corresponding relation between process variable and Complication of dialysis in table 1 dialysis procedure
Neutral net builds module, for building neural network model, including to the selection of network various parameters with establish, model
Foundation, and network model is trained；Using the multilayer feedforward neural network trained by Back Propagation Algorithm, letter below
Claim bp neutral net, the bp neutral net of use has Three Tiered Network Architecture, using regular former piece as bp neutral net input
Layer, input layer has 8 process variables, corresponding 8 nodes；As the output layer of bp neutral net, output layer has 7 to consequent
Complication, corresponding 7 nodes；The establishment of hidden layer neuron number is using formula (1) and formula (2):
Wherein, nimplicitFor hidden layer neuron number, ninputFor input layer number, noutputFor output layer neuron
Number, the value of a is 1～10, herein, determines that node in hidden layer is 14；In described hidden layer, network weight is anti-according to error
Carry out dynamic error correction determination to propagation principle, the training of neutral net adopts quick self-adapted algorithm, using with additional
The gradient descent method traingdx function of activity level method and adjusting learning rate is as the training function of network, e-learning function
It is the gradient descent method function learngdm with momentum term, performance function using sum of squared errors function sse and is set as 0,
Initial learn speed is 0.01, and momentum constant is 0.95, and maximum cycle is 2000 times, and each layer transfer function is takes the logarithm
S type function log-sigmoid；
Rule in expert system knowledge base is table 1 as the input of network and output, completes the training to neutral net；
Monitoring result output module, its function is: by the actual monitoring value input instruction of process variable in dialysis on-line monitoring
The model perfected, obtains the output result of model；This monitoring device corresponding Complication of dialysis output result in the range of 0-1,
The probability size that complication occurs is established by the degree of membership between 0 to 1, the symbolical meaningses of probability degree of membership in complication
As table 3:
Table 3: the degree of membership of probability in complication
2. the hemodialysis on-Line Monitor Device of fusion fuzzy neural network according to claim 1 and regular pattern composite specialist system,
It is characterized in that: realized Bu Tong concurrent with cause to same process variable to accurately portraying of process variable using fuzzy logic
The deep discrimination of disease.
3. the hemodialysis on-Line Monitor Device of fusion fuzzy neural network according to claim 1 and regular pattern composite specialist system,
It is characterized in that: establish fuzzy membership function using expertise knowledge, realize the process of input bp neutral net is become
The Fuzzy processing of value, disclosure satisfy that the needs of bp neutral net input data, need not carry out pretreatment to input data again；
Additionally, the self-learning capability of neutral net makes renewal and the increase of knowledge base, improve the expansion of system, thus realize right
More larger range of complication diagnosis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410202258.XA CN103984861B (en) | 2014-05-14 | 2014-05-14 | Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410202258.XA CN103984861B (en) | 2014-05-14 | 2014-05-14 | Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103984861A CN103984861A (en) | 2014-08-13 |
| CN103984861B true CN103984861B (en) | 2017-01-18 |
Family
ID=51276829
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410202258.XA Expired - Fee Related CN103984861B (en) | 2014-05-14 | 2014-05-14 | Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103984861B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107576703A (en) * | 2017-10-12 | 2018-01-12 | 天津科技大学 | A kind of domestic glucometer based on fuzzy system |
| CN112957553B (en) * | 2021-02-01 | 2022-02-01 | 南通市第二人民医院 | Intelligent monitoring system and method for complications of hemodialysis patients |
| CN113289098B (en) * | 2021-05-25 | 2022-06-03 | 南通市第二人民医院 | Dialysate monitoring method and device based on hemodialysis device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1547149A (en) * | 2003-12-04 | 2004-11-17 | 上海交通大学 | Method for implementing brain glioma computer aided diagnosis system based on data mining |
| CN101793887A (en) * | 2010-01-09 | 2010-08-04 | 中国水产科学研究院黄海水产研究所 | Construction method of fuzzy neural network expert system for water quality assessment in turbot culture |
| CN102737153A (en) * | 2011-04-07 | 2012-10-17 | 中国人民解放军第三〇二医院 | Neural network hepatofibrosis diagnosis method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7326576B2 (en) * | 2003-04-09 | 2008-02-05 | Prescient Medical, Inc. | Raman spectroscopic monitoring of hemodialysis |
-
2014
- 2014-05-14 CN CN201410202258.XA patent/CN103984861B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1547149A (en) * | 2003-12-04 | 2004-11-17 | 上海交通大学 | Method for implementing brain glioma computer aided diagnosis system based on data mining |
| CN101793887A (en) * | 2010-01-09 | 2010-08-04 | 中国水产科学研究院黄海水产研究所 | Construction method of fuzzy neural network expert system for water quality assessment in turbot culture |
| CN102737153A (en) * | 2011-04-07 | 2012-10-17 | 中国人民解放军第三〇二医院 | Neural network hepatofibrosis diagnosis method and system |
Non-Patent Citations (1)
| Title |
|---|
| 基于神经网络和模糊理论的心功能辅助诊断专家系统;谢洪波 等;《天津大学学报》;20000930;第33卷(第5期);第610-614页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103984861A (en) | 2014-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2007344229B2 (en) | Follow-up of the vascular access of a dialyzed patient. | |
| Kline et al. | Emergency clinician–performed compression ultrasonography for deep venous thrombosis of the lower extremity | |
| CN103984861B (en) | Fuzzy neural network and rule-based expert system fused online hemodialysis monitoring device | |
| CN110111884A (en) | A kind of man-machine coordination intelligent medical treatment aid decision-making system based on CMKMC | |
| CN105288767B (en) | The device and method of extracorporeal circulation of blood failure for identification | |
| Anninou et al. | Modeling health diseases using competitive fuzzy cognitive maps | |
| Hussain et al. | Personal home healthcare system for the cardiac patient of smart city using fuzzy logic | |
| Mirza et al. | A fuzzy logic-based system for anaesthesia monitoring | |
| Habib et al. | A neutrosophic clinical decision-making system for cardiovascular diseases risk analysis | |
| Huang et al. | A deep segmentation network of stent structs based on IoT for interventional cardiovascular diagnosis | |
| Fahmy et al. | A decision support healthcare system based on IoT and neural network technique | |
| Lin et al. | Detection of venous needle dislodgement during haemodialysis using fractional order shape index ratio and fuzzy colour relation analysis | |
| CN115985491B (en) | Medical information processing system of interventional ventricular catheter pump | |
| Omari et al. | A new approach for blood pressure estimation based on phonocardiogram | |
| Wu et al. | Peripheral arterial disease screening for hemodialysis patients using a fractional‐order integrator and transition probability decision‐making model | |
| Shih et al. | An early warning system for hemodialysis complications utilizing transfer learning from HD IoT dataset | |
| Devi et al. | Design methodology of a fuzzy knowledgebase system to predict the risk of diabetic nephropathy | |
| Aydın et al. | A new proposal for early stage diagnosis of urinary tract infection using computers aid systems | |
| Rezapour | Forecasting Surgical Outcomes Using a Fuzzy-Based Decision System | |
| CN115359907A (en) | Intelligent diagnosis system and method for screening heart diseases on chest X-ray film | |
| CN112957553B (en) | Intelligent monitoring system and method for complications of hemodialysis patients | |
| Fernández et al. | Artificial neural networks applications in dialysis | |
| CN204815059U (en) | A device for discerning external blood circulation trouble | |
| Ece et al. | Classification of artificial intelligence based coronary artery Stenosis | |
| Santosa et al. | Application of fuzzy logic Sugeno methods for diagnosis typhoid fever disease and dengue hemorrhagic fever |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170118 Termination date: 20190514 |