Disclosure of Invention
The invention provides a similarity matching diagnosis monitoring integrated medical system, which comprises a client, a workstation, a cloud server, an image acquisition device and a physiological parameter sensor, wherein the physiological parameter sensor is transmitted to a network coordinator through a short-distance wireless communication technology to be connected with the client, the client is connected with the cloud server through the internet, the image acquisition device is connected with the workstation and uploads acquired images to the cloud server, a patient user uploads basic information of the patient user to the cloud server through the client, a decision user uploads case general information, main symptoms and signs of the patient user, a test result, pathological information, basic judgment of the patient user's condition, treatment means, a diagnosis result, a diagnosis, The system comprises a post-operation symptom, nursing suggestions and strategies, a cloud server, a diagnosis case base is stored in the cloud server, the cloud server comprehensively forms cases for various uploaded information, a decision user searches the diagnosis case base through a workstation, the most similar case is found through similarity matching in the diagnosis case base, an image acquisition device comprises a CT (computed tomography) machine, an ultrasonic instrument and a nuclear magnetic resonance instrument, physiological parameter sensors comprise heart rate acquisition sensors, a sphygmomanometer, an oximeter and an accelerometer, the wearable device is a wearable device, the workstation performs image fusion on two images of acquired CT images, ultrasonic images and nuclear magnetic resonance images, the fused images are transmitted to the cloud server, and the workstation and the cloud server are communicated through an encryption algorithm.
The decision-making user searches the diagnosis and treatment case library through the workstation, finds the most similar case by matching the similarity in the diagnosis and treatment case library, and comprises the following specific steps:
step 1, defining characteristic attributes of a source decision case according to medical clinical diagnosis and treatment characteristics, and establishing a diagnosis and treatment case library;
step 2, after the patient user visits a doctor, inputting characteristic attribute information through a human-computer interaction interface, and extracting a characteristic vector;
wherein the feature vector includes:
case profile information vectors, which include characteristic attributes for case numbers, case names, time of occurrence of the case, attending physician's name, name of the resident, and name of the caregiver;
a patient basic information vector which comprises characteristic attributes of the sex, the age, the height, the weight, the family history, the health history, the drug allergy history, the admission date and the discharge date of a patient user;
vectors of the patient user's chief symptoms and signs, which include characteristic attributes of the patient user's subjective abnormal sensations, patient user signs, etc.;
test result vectors including characteristic attributes including blood routine, immune combination, coagulogram, urine routine, blood gas analysis, B-mode ultrasonography, CT examination, electrocardio, blood pressure and blood oxygen saturation;
pathological information vectors including the tumor size, the number of the invaded lymph nodes, whether the nodules occur or not, the malignant tumor degree, the position of the tumor mass, the quadrant of the tumor mass, the distribution characteristics and the growth characteristics in the pathological information vectors;
the basic judgment vector of the patient condition of the patient user comprises the subjective judgment result of the decision-making user, namely characteristic attributes including malignancy, benign, easy recurrence and difficult recurrence;
treatment vectors, characteristic attributes including disease description, diagnosis process, whether radiotherapy is applied, treatment process, treatment effect and intermediate examination;
a post-operative symptom vector, i.e. a characteristic attribute including a description of the user's symptoms of the breast cancer patient after treatment;
vectors of care recommendations and strategies for patient-user specific illness and symptom characteristics.
Step 3, inputting the characteristic vector, and finding out a relevant case in the diagnosis and treatment case library by a fuzzy diagnosis and treatment knowledge finding program;
step 4, acquiring the optimal weight of each characteristic attribute in each characteristic vector through a genetic algorithm; the method specifically comprises the following steps:
step 4.1, encoding the characteristic attribute;
step 4.2, generating initial weights of the characteristic attributes;
step 4.3, calculating fitness and keeping the preferred chromosome;
step 4.4, finding an accurate value;
4.5, copying and selecting by using a roulette selection method;
step 4.6, crossing;
step 4.7, mutation;
step 4.8, preferably chromosomes;
step 4.9, replacing the existing chromosome;
step 4.10, whether the evolution iteration number reaches the upper limit, if so, stopping iteration, entering step 11, otherwise, returning to step 3, and continuing iteration;
step 4.11, inversely coding the current chromosome to generate an optimal weight;
and 5, searching in a diagnosis and treatment case library through a case search program by combining the optimal weight and each characteristic attribute, and analyzing the similarity to obtain the most similar case, wherein the method specifically comprises the following steps:
step 5.1, determining positive and negative ideal cases of decision users, constructing a fuzzy matrix and carrying out normalization;
step 5.2, carrying out weighted calculation on the fuzzy matrix;
step 5.3, weighting the index values of each attribute of the positive and negative ideal cases;
step 5.4, calculating the distance between the positive and negative ideal cases, the target case and the input case,
wherein i is a feature vector ordinal number; j is a feature attribute ordinal; n is the total number of the feature vectors; l is the total number of the characteristic attributes; x is the number ofijThe j characteristic attribute value of the i characteristic vector;is the distance to the positive ideal point; IDjIs the distance to the negative ideal point; IDTThe distance between each case and the target case;in order to be able to take the point value as ideal,is a negative ideal point value, xTFor target point values, d (,) is a distance function between fuzzy numbers;
step 5.5, calculating the degree of closeness, determining the sequence among the cases, finding the case most similar to the target case, calculating the degree of closeness between the cases and the target case,
the TH is the fitting degree, the higher the TH value is, the more the two cases are close, and the case with the highest fitting degree is selected as the most similar case;
step 6, returning the most similar cases to the decision user, if the requirements are met, generating diagnosis cases, and entering step 7; if the requirements are not met, the decision-making user inputs specific requirements, a case correction program is started, the obtained cases are corrected until the cases meet the requirements, and the step 7 is carried out;
and 7: and (4) the decision user reviews and evaluates the value of the retrieved case, if the case is considered to be valuable, a corresponding diagnosis conclusion is generated and stored in the diagnosis and treatment case library, and if the case is considered to be not valuable, a corresponding diagnosis conclusion is generated and quit.
The invention has the beneficial effects that:
(1) the case closest to the patient is searched through similarity matching, so that a reliable basis treatment scheme and a basis for selecting a patient treatment environment are provided for diagnosis of a doctor, and the labor intensity of the doctor is greatly reduced.
(2) The detail display of the focus image of the patient is realized through image fusion, and a powerful basis is provided for the diagnosis of a doctor;
(3) the posture of the patient is recognized through the judgment of the motion state, and the state of the patient is monitored in real time;
(4) the privacy of the patient is effectively guaranteed not to be disclosed through encrypted communication.
Detailed Description
The invention is further described with reference to the following figures and examples.
Embodiments of the present invention are illustrated with reference to fig. 1-2.
A diagnosis and monitoring integrated medical system matched with similarity comprises a client, a workstation, a cloud server, an image acquisition device and a physiological parameter sensor, wherein the physiological parameter sensor is transmitted to a network coordinator through a short-distance wireless communication technology to be connected with the client, the client is connected with the cloud server through the Internet, the image acquisition device is connected with the workstation to upload acquired images to the cloud server, a patient user uploads basic information of the patient user to the cloud server through the client, a decision user uploads case profile information, main symptoms and signs of the patient user, a test result, pathological information, basic judgment of patient condition, treatment means, postoperative symptoms, nursing suggestions and strategies to the cloud server through the workstation, the cloud server stores a diagnosis and treatment case library and comprehensively forms cases for various uploaded information, the decision-making user searches the diagnosis and treatment case base through the workstation, finds the most similar case by matching the similarity in the diagnosis and treatment case base, the image acquisition device comprises a CT machine, an ultrasonic instrument and a nuclear magnetic resonance instrument, the physiological parameter sensor comprises a heart rate acquisition sensor, a sphygmomanometer, an oximeter and an accelerometer, the wearable device is a wearable device, the workstation performs image fusion on two images of the acquired CT image, the ultrasonic image and the nuclear magnetic resonance image, the fused image is uploaded to a cloud server, and the workstation communicates with the cloud server by using an encryption algorithm.
The decision-making user searches the diagnosis and treatment case library through the workstation, finds the most similar case by matching the similarity in the diagnosis and treatment case library, and comprises the following specific steps:
step 1, defining characteristic attributes of a source decision case according to medical clinical diagnosis and treatment characteristics, and establishing a diagnosis and treatment case library;
step 2, after the patient user visits a doctor, inputting characteristic attribute information through a human-computer interaction interface, and extracting a characteristic vector;
wherein the feature vector includes:
case profile information vectors, which include characteristic attributes for case numbers, case names, time of occurrence of the case, attending physician's name, name of the resident, and name of the caregiver;
a patient basic information vector which comprises characteristic attributes of the sex, the age, the height, the weight, the family history, the health history, the drug allergy history, the admission date and the discharge date of a patient user;
vectors of the patient user's chief symptoms and signs, which include characteristic attributes of the patient user's subjective abnormal sensations, patient user signs, etc.;
test result vectors including characteristic attributes including blood routine, immune combination, coagulogram, urine routine, blood gas analysis, B-mode ultrasonography, CT examination, electrocardio, blood pressure and blood oxygen saturation;
pathological information vectors including the tumor size, the number of the invaded lymph nodes, whether the nodules occur or not, the malignant tumor degree, the position of the tumor mass, the quadrant of the tumor mass, the distribution characteristics and the growth characteristics in the pathological information vectors;
the basic judgment vector of the patient condition of the patient user comprises the subjective judgment result of the decision-making user, namely characteristic attributes including malignancy, benign, easy recurrence and difficult recurrence;
treatment vectors, characteristic attributes including disease description, diagnosis process, whether radiotherapy is applied, treatment process, treatment effect and intermediate examination;
a post-operative symptom vector, i.e. a characteristic attribute including a description of the user's symptoms of the breast cancer patient after treatment;
vectors of care recommendations and strategies for patient-user specific illness and symptom characteristics.
Step 3, inputting the characteristic vector, and finding out a relevant case in the diagnosis and treatment case library by a fuzzy diagnosis and treatment knowledge finding program;
step 4, acquiring the optimal weight of each characteristic attribute in each characteristic vector through a genetic algorithm; the method specifically comprises the following steps:
step 4.1, encoding the characteristic attribute;
step 4.2, generating initial weights of the characteristic attributes;
step 4.3, calculating fitness and keeping the preferred chromosome;
step 4.4, finding an accurate value;
4.5, copying and selecting by using a roulette selection method;
step 4.6, crossing;
step 4.7, mutation;
step 4.8, preferably chromosomes;
step 4.9, replacing the existing chromosome;
step 4.10, whether the evolution iteration number reaches the upper limit, if so, stopping iteration, entering step 11, otherwise, returning to step 3, and continuing iteration;
step 4.11, inversely coding the current chromosome to generate an optimal weight;
and 5, searching in a diagnosis and treatment case library through a case search program by combining the optimal weight and each characteristic attribute, and analyzing the similarity to obtain the most similar case, wherein the method specifically comprises the following steps:
step 5.1, determining positive and negative ideal cases of decision users, constructing a fuzzy matrix and carrying out normalization;
step 5.2, carrying out weighted calculation on the fuzzy matrix;
step 5.3, weighting the index values of each attribute of the positive and negative ideal cases;
step 5.4, calculating the distance between the positive and negative ideal cases, the target case and the input case,
wherein i is a feature vector ordinal number; j is a feature attribute ordinal; n is the total number of the feature vectors; l is the total number of the characteristic attributes; x is the number ofijThe j characteristic attribute value of the i characteristic vector;is the distance to the positive ideal point;is the distance to the negative ideal point; IDTThe distance between each case and the target case;in order to be able to take the point value as ideal,is a negative ideal point value, xTFor target point values, d (,) is a distance function between fuzzy numbers;
step 5.5, calculating the degree of closeness, determining the sequence among the cases, finding the case most similar to the target case, calculating the degree of closeness between the cases and the target case,
the TH is the fitting degree, the higher the TH value is, the more the two cases are close, and the case with the highest fitting degree is selected as the most similar case;
step 6, returning the most similar cases to the decision user, if the requirements are met, generating diagnosis cases, and entering step 7; if the requirements are not met, the decision-making user inputs specific requirements, a case correction program is started, the obtained cases are corrected until the cases meet the requirements, and the step 7 is carried out;
and 7: and (4) the decision user reviews and evaluates the value of the retrieved case, if the case is considered to be valuable, a corresponding diagnosis conclusion is generated and stored in the diagnosis and treatment case library, and if the case is considered to be not valuable, a corresponding diagnosis conclusion is generated and quit.
The workstation and the cloud server communicate by using an encryption algorithm, and the encryption and decryption processes are as follows:
step 1, a sender shares a secret key with a receiver, and a shared secret key is generated according to a public key of the receiver;
step 2, the sender uses the shared secret key and other selectable parameters as the input of a secret key derivation function, and the derivation result forms a symmetric encryption secret key and an MAC secret key pair;
step 3, using a symmetric encryption algorithm to take a symmetric encryption key and a plaintext as input to generate a ciphertext;
step 4, using the encryption information, the MAC key and other public parameters as input, and selecting an MAC function to generate an identifier;
step 5, the sender uses the temporary public key, the identifier and the ciphertext to form encrypted information and sends the encrypted information to the receiver;
step 6, the receiver receives the encrypted information and obtains the temporary public key, the identifier and the ciphertext of the sender, so that the encrypted information is decomposed;
step 7, the receiver multiplies the received temporary public key by the private key of the receiver;
step 8, using the shared key and the public parameters used by other senders, the receiver performs the same derivation process as the sender, thereby obtaining an encryption key and an MAC key pair;
step 9, using the encrypted information, the MAC key and other public parameters used by the sender as input, calculating a new identifier by the receiver, comparing the new identifier with the received identifier, and if the new identifier is different from the received identifier, failing to verify and refusing to receive the ciphertext;
and step 10, if the new identifier is the same as the received identifier, the receiver continues the decryption process, and outputs the plaintext of the sender by using the symmetric decryption algorithm and taking the encrypted information and the decryption key as outputs.
The workstation or the cloud server serves as a sender, and the cloud server or the workstation serves as a receiver.
Wherein, physiological parameter sensor is wearable equipment, and it judges patient's motion state through the acceleration data of accelerometer, and specific judgement process is as follows:
step 1, acquiring acceleration data from an accelerometer;
and step 2, pre-processing the data,
the method comprises the steps of generating the acquired acceleration data into acceleration time sequence data, windowing time domain of the acceleration time sequence data by adopting a sliding window, processing and identifying the acceleration time sequence data obtained by each windowing, segmenting and extracting sequences by using a time axis by the sliding window, and defining a window length w and an overlapping window length o.
For a time series of accelerations x1,x2...xn.., the first window is x1,x2...xwThe second window is { x }w-o,xw-o+1...x2w-o-1The third window is { x }2w-2o-1,x2w-2o...x3w-2o-2N, each window is { x }n(w-o)-n+1,xn(w-o)-n+2...xn(w-o)+w-nWhere w is 60 and o is 10. x is the number ofnAcceleration time series data.
The overlapping has the effects of reducing the frequency spectrum leakage and reducing the damage of signal truncation to information.
And 3, generating the data characteristics,
the acceleration data features include time domain features including an accelerated mean, root mean square, variance, and absolute mean over a window length w, and frequency domain features that are energy over the window length w.
Mean valueThe average magnitude of the acceleration value is the accelerated direct current component, and the calculation formula is as follows:
the root mean square D is the total size of the acceleration values, and the calculation formula is as follows:
variance σxThe calculation formula is as follows:
the variance clearly reflects the intensity of the data change, wherein-Is the average of the accelerations.
The absolute mean difference K is calculated as:
the calculation method of the energy E is that after FFT is carried out on the signal, the sum of the squares of the amplitudes of all the components is as follows:
Fithe amplitude of the ith component after the acceleration time FFT is shown, and N is the number of the components.
And respectively obtaining an average value sequence, a root mean square sequence, a variance sequence, an absolute average difference sequence and an energy sequence of a plurality of windows through the calculation of the average value, the root mean square, the variance, the absolute average difference and the energy.
Feature generation is crucial in the task of pattern recognition, given a set of measures, the goal of feature generation is to explore the natural patterns of the acquired data and the re-representation of the information. A good feature generation process can compress basic information of data and simultaneously eliminate redundant information, so that the dimension of an original data space is reduced.
Step 4, the matching identification is carried out,
generating a sample sequence, wherein the sample sequence comprises an average value sequence, a root mean square sequence, a variance sequence, an absolute average difference sequence and an energy sequence in a walking state, a running state, a jumping state, a falling state and a static state, and the sample sequence is preset or is set by a user by executing a corresponding state;
and respectively matching the obtained average value sequence, root mean square sequence, variance sequence, absolute average difference sequence and energy sequence with sample sequences in different states, and taking the state of the sample sequence with the highest matching degree as the motion state of the current user.
The method comprises the following steps of generating a focused color image by using an ultrasonic image and a nuclear magnetic resonance image, wherein the fusion process is as follows:
step 1, converting an image A, B from an RGB space to an HSV space through HSV conversion, and obtaining H, S, V three components;
step 2, obtaining a fused H component through neighborhood superposition;
step 3, obtaining a fused S component and a fused V component through gradient calculation;
and 4, forming a final fusion image through HSV inverse transformation according to the H component, the S component and the V component in the step.
Wherein, the step 2 specifically comprises the following steps:
step 2.1, initialization, n ═ 1, Yi,j(0)=0,Li,j(0)=1,θi,j(0)=1,
Step 2.2, the iteration,
wherein n is the number of iterations;linear link input for image A, B after nth superposition at point (i, j);the liveness of the images A and B after the nth superposition at the point (i, j) is shown;for image a, B the H component after the nth superposition at point (i, j);dynamic threshold for image a, B after nth overlay at point (i, j);outputting the binary output of the image A and B after the nth superposition at the point (i, j);the high binary times of the images A and B after the nth superposition at the point (i, j) are obtained; alpha is a time decay constant; beta is the link strength coefficient; wijA weight coefficient at point (i, j); vL、VθIs the amplification factor; k, l are points (i, j) which are vertically and horizontally adjacent link extents.
Step 2.3, judging that N is larger than N, wherein N is an iteration threshold; entering step 4, otherwise, returning to step 2;
step 2.4, counting, respectively calculating the total times of high binary values of the image A and the image B,
step 2.5, perform the above steps 1-4 on all points (i, j) in image A and image B to obtain TA(i, j) and TB(i,j);
And 2.6, fusing the components,
wherein HAB(i, j) is the H component after fusion at point (i, j), HA(i,j)、hB(i, j) is the H component, T, of the original image A, B at point (i, j)A(i,j)、TB(i, j) is the total number of times the image A, B was high at point (i, j),
wherein, the step 3 is specifically as follows:
step 3.1, gradient calculation
Wherein the content of the first and second substances,
wherein the content of the first and second substances,is the gradient, G, of the original image A, B at point (i, j)i()、Gj()、Gij()、Gji() Is a function of the transverse, vertical, 45 DEG oblique, 135 DEG oblique variation at point (i, j), sA(i,j)、sB(i, j) is the S component of the original image A, B at point (i, j),
in addition, expressions that are not avoided to be too long cannot be shown, and { } here denotes the covering of long expressions, i.e.
And 3.2, fusing the components,
SAB(i, j) is the S component after fusion at point (i, j);
step 3.3, obtaining the V component V of the fused image A, B at point (i, j) through the above steps 3.1-3.2AB(i,j)。
Wherein, the image A is an ultrasonic image or an NMR image, and the image B is an NMR image or an ultrasonic image.
The above-described embodiment merely represents one embodiment of the present invention, but is not to be construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.