TWI827435B - A urination detection method and system - Google Patents

Abstract

The present invention provides a urination detection method and system, which acquires a urination image information by a capture unit, and transmits the urination test result to a cloud database through an identification module, so as to record the urination situation of the user in daily life in detail, and provide the real-time urination health assessment, so that a physician can receive the patient's urination information in real time.

Description

Urine detection method and system

A urination health assessment method, in particular, relates to a urination detection method and its system.

According to statistics from the Ministry of Health and Welfare, about 10% of men over 70 years old suffer from acute urinary retention, and more than 33% of men over 80 years old suffer from acute urinary retention. Moreover, according to the National Health Insurance Medical Statistics Annual Report in 2011, urinary system diseases started in the 40s. -The number of medical consultations for 49-year-olds is about 330,000, the number of medical consultations for 50-59-year-olds is about 390,000, and the number of medical consultations for 60-69-year-olds is as high as about 450,000, among which urinary system The disease is often associated with symptoms of the lower urinary tract system, such as frequent urination, urgent urination, difficulty in urinating, urinary incontinence, and urinary retention, which affects the quality of life. It is a serious public health problem and awaits the development of a non-invasive urine detection method. Differential diagnosis.

The current method of diagnosis is for urologists to diagnose patients' urinary status by regularly asking patients to return for consultation and then taking questionnaires, such as: IPSS (International Prostate Symptom Score), VAUS (Visual Analogue Uroflowmetry Score), combined with urine flow rate examination. (Uroflowmetry) is used to diagnose urinary health, but its questionnaire method cannot objectively indicate the actual urinary status. At the same time, it is actually difficult for patients to remember their urinary status throughout the month, which makes the questionnaire method often a mere formality.

Uroflowmetry is a continuous urinary flow rate curve that uses a urinary flow recorder to depict the urination process. It can obtain information such as urinary flow rate graphics, urinary volume, urinary flow time, and urinary flow rate (maximum, average). This type of examination can only allow patients to urinate in the hospital, and then analyze the urine flow rate and urine flow rate obtained by it. It is limited by the examination location and time. Moreover, the current urine flow velocity examination can only be measured after urinating in a specific container. , if it is difficult for patients with limited mobility to cooperate with the examination, therefore, how to enable patients to record their urination situations in daily life in detail and conduct immediate urinary health assessment is a very important issue.

The main purpose of the present invention is to provide a urination detection method that calculates the coordinate values of urination image information to generate urination detection results and greatly improve the accuracy of the examination.

Another object of the present invention is to provide a urination detection system that installs an acquisition unit in the bathroom and toilet to obtain urination image information, and transmits the urination detection results to a cloud database through the recognition module, so that doctors can obtain real-time information. Obtain patient urination information, not limited to testing location and time.

In order to achieve the above object, one embodiment of the present invention discloses a urination detection method. The steps include: acquiring a urination image information; marking a position of the urination image information, and obtaining a plurality of coordinate information corresponding to the position; calculating The coordinate information generates urination angle information and urination distance information; and an algorithm is used to calculate the urination angle information and the urination distance information to generate a urination detection result.

In a preferred embodiment, in the step of marking a position of the urination image information and obtaining a plurality of coordinate information corresponding to the position, the position includes a human body joint point and a urination trajectory, and the coordinate information includes a The coordinates of human body joint points and the coordinates of a urination trajectory.

In a preferred embodiment, in the step of using an algorithm to calculate the urination angle information and the urination distance information to generate a urination detection result, a training model is established using a tachometer information, and the training model uses the algorithm The urination angle information and the urination distance information are calculated to generate the urination detection result.

In a preferred embodiment, the algorithm is a Hidden Markov Model (HMM).

In a preferred embodiment, in the step of marking a position of the urination image information and obtaining a plurality of coordinate information corresponding to the position, a plurality of human body image information is input for deep learning to generate a thermal image recognition information, based on The urination image information is compared with the thermal image identification information to mark the position in the urination image information.

In order to achieve another of the above objectives, one embodiment of the present invention discloses a urination detection system, which includes: a capture unit to capture urination image information; and an identification module that is connected to the capture unit and a cloud respectively. The database signal connection performs calculations based on the urination image information, generates and transmits a urination test result to the cloud database.

In a preferred embodiment, the identification module marks a position of the urination image information based on the urination image information, obtains and performs calculations based on a plurality of coordinate information corresponding to the position, and generates urination angle information and urination distance information. , and generate the urination test result based on the urination angle information and the urination distance information.

In a preferred embodiment, the positioning includes a human body joint point and a urination trajectory, and the coordinate information includes a human body joint point coordinate value and a urination trajectory coordinate value.

In a preferred embodiment, an automatic lifting device is included, which is signal-connected to the identification module. The automatic lifting device carries the capturing unit, and adjusts the height of the automatic lifting device according to the coordinate values of the human body joint points.

In a preferred embodiment, the capture unit is an infrared thermal imager.

The beneficial effect of the present invention is that it does not need to be limited to the detection location, and can be set according to the location of the bathroom used by different users. Through the obtained urination image information, it can automatically calculate and generate urination detection results to improve detection accuracy.

1: Capture unit

2:Identification module

3: Automatic lifting device

4: Cloud database

D: Urinating distance information

J0: nose

J1:Left eye

J2: Right eye

J3: Left ear

J4: Right ear

J5:Left shoulder

J6: Right shoulder

J7:Left elbow

J8: Right elbow

J9:Left wrist

J10: Right wrist

J11: Left hip

J12: Right hip

J13:Left knee

J14: Right knee

J15:Left ankle

J16: Right ankle

M: coordinates

R: final coordinate point

S: starting point of urination

S1: Steps

S2: Step

S3: Steps

S4: Steps

V _L : vertical line

θ: urination angle information

The first picture: It is a system schematic diagram of an embodiment of the present invention; The second picture: It is a method flow chart of an embodiment of the present invention; The third picture A: It is a human body joint point of an embodiment of the present invention Schematic diagram; Figure 3B: It is a schematic diagram of the operation of one embodiment of the present invention; Figure 4A: It is a diagram of the urination detection experiment results of subject 1 according to one embodiment of the present invention; Figure 4B: It is It is a picture of the urination detection experiment results of subject 2 according to one embodiment of the present invention; the fourth picture C: it is the result picture of the urination detection experiment of subject 3 according to one embodiment of the present invention; the fourth picture D: it is It is a picture of the urination test results of subject 4 according to one embodiment of the present invention; the fourth picture E: it is a picture of the results of the urination test of subject 5 according to one embodiment of the present invention; the fourth picture F: it is It is a picture of the urination detection experiment results of subject 6 according to one embodiment of the present invention; the fourth picture G: it is a picture of the urination detection experiment results of subject 7 according to one embodiment of the present invention; the fourth picture H: it is This is a diagram showing the results of the urination detection experiment of subject 8 according to one embodiment of the present invention; Figure 5A: It is the Uroflowmetry result graph of subject 1 according to one embodiment of the present invention; Figure 5B: It is the Uroflowmetry result graph of subject 2 according to one embodiment of the present invention; Figure 5C : It is the Uroflowmetry result diagram of subject 3 according to one embodiment of the present invention; The fifth D diagram is the Uroflowmetry result diagram of subject 4 according to one embodiment of the present invention; The fifth E diagram is the Uroflowmetry result diagram of subject 3 according to one embodiment of the present invention; The Uroflowmetry result chart of subject 5 according to one embodiment of the invention; the fifth F chart: it is the Uroflowmetry result chart of subject 6 according to one embodiment of the present invention; the fifth G chart: it is one implementation of the present invention. For example, the Uroflowmetry result chart of subject 7; Figure 5H: It is the Uroflowmetry result chart of subject 8 according to one embodiment of the present invention; Figure 5I: It is the Uroflowmetry result chart of subject 8 according to one embodiment of the present invention. Figure 9 is a Uroflowmetry result graph; Figure 5J is a Uroflowmetry result graph of subject 10 according to one embodiment of the present invention; Figure 5K is a Uroflowmetry result graph for subject 11 according to one embodiment of the present invention. The result graph; the fifth L graph: it is the Uroflowmetry result graph of the subject 12 in one embodiment of the present invention; the fifth M graph: it is the Uroflowmetry result graph of the subject 13 in one embodiment of the present invention; Figure 5N: It is the Uroflowmetry result graph of the subject 14 in one embodiment of the present invention; Figure 5O: It is the Uroflowmetry result graph of the subject 15 in one embodiment of the present invention; and Figure 5P : This is the Uroflowmetry result chart of subject 16 according to one embodiment of the present invention.

In order to make the above and/or other objects, effects, and features of the present invention more obvious and understandable, the following is a detailed description of the preferred embodiments:

Please refer to the first figure, which is a system schematic diagram of an embodiment of the present invention. As shown in the figure, a urination detection system according to one embodiment of the present invention includes: a capture unit 1, an identification module 2, an automatic lifting unit Device 3 and cloud database 4, identification module 2 are connected to the acquisition unit 1, automatic lifting device 3 and cloud database 4 respectively, and their operation methods are described in detail as follows:

The capture unit 1 includes but is not limited to an infrared thermal imager. In one embodiment, the capture unit 1 is installed in the bathroom used by the user to capture urination image information when the user urinates, but this is not limited to this. .

The recognition module 2 includes but is not limited to an AI edge computing platform. In one embodiment, the recognition module 2 annotates the position of the urination image information based on the urination image information, where the positioning includes human body joint points and urination trajectories, and obtains the corresponding position. A plurality of coordinate information, wherein the coordinate information includes human body joint point coordinate values and urination trajectory coordinate values, and the coordinate information is calculated to generate a urination detection result. In one embodiment, the urination detection result is Abnormal urination means that there is a possibility of acute urinary retention at any time, but this is not the case.

The automatic lifting device 3 carries the capturing unit 1 and adjusts the height of the automatic lifting device according to the coordinates of the joint points of the human body. In one embodiment, under the best circumstances, the capturing unit 1 is set parallel to the waist of the human body. Therefore, The capture unit 1 can be loaded on the automatic lifting device 3, and after capturing the image information, the position of the capture unit 1 can be adjusted to the most suitable height for the human body based on the coordinate points of the joint points of the human body, but this is not limited to this.

Furthermore, the urination test results can be uploaded to the cloud database 4. At this time, the medical staff can connect to the cloud database 4 to view the patient's urination test results. Once an abnormality occurs, the patient can be notified to a medical institution for treatment. to avoid worsening of urinary retention.

Please also refer to the second figure, which is a method flow chart according to an embodiment of the present invention. As shown in the figure, the steps of the urination detection method according to one embodiment of the present invention are as follows: Step S1: Acquire urination image information; Step S2: Mark a position of the urination image information, and obtain a plurality of coordinate information corresponding to the position; Step S3: Calculate the coordinate information to generate a urination angle information and a urination distance information; and Step S4: Use a The algorithm calculates the urination angle information and the urination distance information to generate a urination detection result.

As shown in step S1, the urination image information is captured by the capture unit 1. When the capture unit 1 is an infrared thermal imager, the urination image information is presented in the form of a thermal image. The infrared thermal imager converts radiation energy into As an electrical signal, a variety of different colors are used to display the distribution of different temperatures, so that the entire temperature distribution state is displayed as a visible image. This also has privacy issues, but is not limited to this.

As shown in step S2, when the urination image information of the human body is obtained, the acquisition unit 1 transmits the urination image information to the identification module 2. After receiving the urination image information, the position of the urination image information is correspondingly marked. This positioning includes but It is not limited to human body joint points and urination trajectories. The human body joint points include seventeen joint points as an example. Refer to Figure 3A, which is a schematic diagram of human body joint points according to an embodiment of the present invention. As shown in the figure, nose J0, left eye J1, right eye J2, left ear J3, right ear J4, left shoulder J5, right shoulder J6, left elbow J7, right elbow J8, left wrist J9, right wrist J10, left hip J11, right hip J12, left knee J13, right knee J14, left ankle J15, right ankle J16, and label the joint points at these positions and give coordinate values. But this is not the case.

Furthermore, in one embodiment, a plurality of human body image information will be input for deep learning. The human body image information can be presented in the form of thermal images. The deep learning method, for example, uses data augmentation to expand the training data. set, which is scaled randomly (Random scale), Random rotation and random crop increase the number of samples in the original data set by 4 times to generate a mixed data set, and then import the mixed data set into an RGB image suitable for identifying the human torso. The Open Pose module re-trains the module to generate thermal image recognition information, that is, the thermal image torso recognition module can recognize the thermal image torso. At this time, it performs the process based on the received urination image information and the generated thermal image recognition information. Comparison is used to facilitate annotation and positioning of urination image information and to obtain multiple coordinate information corresponding to the positioning, but this is not limited to this.

As shown in step S3, in one embodiment, please also refer to Figure 3B, which is a schematic diagram of an operation of one embodiment of the present invention. As shown in the figure, the urinary flow recognition algorithm is used to calculate these coordinate information to calculate the urination angle information θ and urination distance information D. First, a vertical line V _L is given in the middle of the urination image, and the urination image is divided into In the left area and the right area, the final coordinate point R (X _R , Y _R ) where urine enters the toilet in the left area is the position with the highest temperature in the image and the minimum value on the Y coordinate axis, continuing in the right area, Use the value of the X coordinate axis of the left wrist J9 and the value of the Y coordinate axis of the left hip J11 as the starting point S (X _J9 , Y _J11 ), and then compare the value of the X coordinate axis of the left wrist J9 with the value of the final coordinate point R The value of the Y coordinate axis is defined as the coordinate M (X _J9 , Y _R ), where the urination distance information D and urination angle information θ are calculated as follows:

Through formula (1)(2)(3), the user's urination distance information D and urination angle information θ can be obtained.

Furthermore, the coordinate values of the urination thickness can also be obtained through the above method, and the urination thickness value can be obtained to provide a more accurate detection mode, but this is not limited to this.

As shown in step S4, the urination distance information D and the urination angle information θ are further calculated using an algorithm to generate a urination detection result. In one embodiment, the algorithm may use a Hidden Markov Model (HMM). , which is a statistical model, assuming that the data set is in the process of multiple tests, and n urination data can be captured each time as training sample parameters, then the training sample parameter observation of the movement acceleration of the data set in each urination process can be obtained Sequence O, formula (4) is as follows: O(m,n)=(O[0][0]O[0][1]…O[0][n]),(O[1][0]O [1][1]…O[1][n]),…,(O[m][0]O[m][1]…O[m][n])…(4)

Then through the hidden Markov model, the conditional probability of the HMM initial model of sequence O can be obtained as P = (O | λ), where λ = {A, B, π} is the HMM initial model parameter, and A is the transition probability (Transition Probability) ; B is the emission probability (Emission Probability); π is the initial probability (Initial Probability)). The HMM initial model parameters can also be imported into the Uroflowmetry results, that is, a training model is established through urinary tachometer information to make the prediction model more accurate.

，以鮑姆-韋爾奇(Baum-Welch)演算法，重新計算初始模型參數即可訓練，並得到序列O的HMM預測模型參數

，最終以Forward-Backward演算法來降低時間複雜度至O(N²T)，藉以求得HMM預測模型參數條件下之條件機率

，如公式(5)，即資料集的測試樣本參數觀察序列

會與訓練樣本參數觀察序列O之機率相等。 In order to obtain the HMM prediction model parameters

, use the Baum-Welch algorithm to recalculate the initial model parameters to train, and obtain the HMM prediction model parameters of sequence O

, and finally use the Forward-Backward algorithm to reduce the time complexity to O(N ² T), thereby obtaining the conditional probability under the parameters of the HMM prediction model.

, such as formula (5), that is, the test sample parameter observation sequence of the data set

It will be equal to the probability of observing sequence O with the training sample parameters.

That is, the HMM prediction model uses the urination data measured by Uroflowmetry and the results of physician diagnosis to find the corresponding feature points in the urination distance information D and urination angle information θ, and distributes the weights of these feature points to obtain each The probability of urination test results and find the highest probability of urination Urine test results, for example: the probability of normal urination is 12% and the probability of abnormal urination is 88%. If the probability of normal urination is less than the probability of abnormal urination, then the urine test result is abnormal urination, but this is not the limit.

In order to more clearly verify the beneficial effects of the test results of the urination test results of one embodiment of the present invention, please refer to the fourth AH diagram and the fifth AP diagram, which are the urination tests of subjects 1-8 according to one embodiment of the present invention. Experimental results and Uroflowmetry results for subjects 1-16. As shown in the figure, the vertical axis is the urination distance information D and the urination angle information θ respectively, and the horizontal axis is time, which is based on the urination distance information D and urination angle information obtained by different subjects in a single urination situation. θ. At the same time, the Uroflowmetry results correspond to the urine flow rate data and urination test results obtained by the same subject in a single urination situation. Table 1 below shows the Uroflowmetry results and HMM prediction model results:

As can be seen from the foregoing, the Uroflowmetry results are the same as the results of the HMM prediction model of one embodiment of the present invention. The classification results show that the urination test results can be effectively obtained with the urination detection method of one embodiment of the present invention, that is, the detection provided by it is verified. The method can be used as clinical reference data.

Table 2 below shows the detailed data of Uroflowmetry

It can be seen from the detailed data of Uroflowmetry that the Uroflowmetry results are based on at least one characteristic point corresponding to the urination test result. For example, the characteristic points are: maximum urine flow rate <15ml/s, urination time >30sec, and further, it can also be It is the proportion of urination time, or the proportion of maximum flow time. Preferably, these feature points can have different weight distribution, or specific feature points can be used as the first priority to filter, so as to obtain more accurate urination detection. result.

In one embodiment, the maximum urine flow rate <15 ml/s is used as the first priority screening, that is, when the maximum urine flow rate is <15 ml/s, it is judged as abnormal urination, and the urination time >30 sec can be used as the criterion. When urinating When the time is >30sec and the maximum urine flow rate is <15ml/s, it is judged as abnormal urination, but this is not the limit.

The aforementioned urination detection results obtained from detailed data of Uroflowmetry are introduced into the urination detection method of this embodiment, and the urination distance information D and urination angle information θ can be further calculated through the HMM prediction model results to generate urination detection results. , to achieve home testing and upload this information to the cloud database 4, so that medical institutions can grasp the information of the subjects in real time, and before the patients develop symptoms of urinary retention, they can be notified to the medical institutions as early as possible treatment.

To sum up, the present invention provides a urination detection method and system. It only needs to use the acquisition unit to obtain urination image information, identify and calculate the urination image information, and generate urination detection results. The detection process is very simple and can be configured. In order to truly present the actual urination situation in the bath and toilet that one is accustomed to, the detection accuracy is improved, and the purpose of the present invention is achieved.

However, the above are only preferred embodiments of the present invention, but they cannot be used to limit the scope of the present invention; therefore, any simple equivalent changes and modifications made based on the patent scope of the present invention and the content of the specification, All are still within the scope of the patent of this invention.

1: Capture unit

2:Identification module

3: Automatic lifting device

4: Cloud database

Claims (10)

A urination detection method, the steps include: Retrieve a urination image information; Mark a position of the urination image information and obtain multiple coordinate information corresponding to the position; Calculate the coordinate information to generate urination angle information and urination distance information; and An algorithm is used to calculate the urination angle information and the urination distance information to generate a urination detection result. According to the urination detection method described in claim 1, in the step of marking a position of the urination image information and obtaining a plurality of coordinate information corresponding to the position, the position includes a human body joint point and a urination trajectory. The coordinate information includes a human body joint point coordinate value and a urination trajectory coordinate value. According to the urination detection method described in claim 1, in the step of calculating the urination angle information and the urination distance information with an algorithm to generate a urination detection result, a training model is established using a tachometer information. The training model The algorithm is used to calculate the urination angle information and the urination distance information to generate the urination detection result. The urination detection method according to claim 1, wherein the algorithm is a Hidden Markov Model (HMM). According to the urination detection method described in claim 1, in the step of marking a position of the urination image information and obtaining a plurality of coordinate information corresponding to the position, a plurality of human body image information is input for deep learning to generate a thermal image Identification information is compared with the urination image information and the thermal image identification information to mark the position in the urination image information. A urination detection system including: An acquisition unit acquires urination image information; and An identification module is respectively connected with the acquisition unit and a cloud database signal, performs calculations based on the urination image information, generates and transmits a urination test result to the cloud database. The urination detection system according to claim 6, wherein the identification module marks a position of the urination image information based on the urination image information, obtains and performs calculations based on a plurality of coordinate information corresponding to the position, and generates a urination angle information and a urination distance information, and generate the urination test result based on the urination angle information and the urination distance information. According to the urination detection system described in claim 7, the positioning includes a human body joint point and a urination trajectory, and the coordinate information includes a human body joint point coordinate value and a urination trajectory coordinate value. The urination detection system according to claim 8 includes an automatic lifting device connected to the identification module signal. The automatic lifting device carries the acquisition unit and adjusts the automatic lifting device according to the coordinate values of the human body joint points. One height. The urination detection system according to claim 6, wherein the capture unit is an infrared thermal imager.