A physiological information monitoring and early warning method and system

Abstract

The present invention discloses a physiological information monitoring and early warning method and system. According to the longitudinal size change and position change of multiple resized images, the characteristic information is transferred to one image, and it is convenient to extract the features later. The position sequence is arranged according to the relationship of the physiological information, and the relationship of the physiological information is placed in the middle, and the related one is at the boundary below and the other is above the middle, so that similar physiological information can be in two positions without losing information during feature extraction. According to different types of values, the size of the time physiological image is adjusted, so that when multiple time physiological images are combined into one adjusted image, it is more accurate to extract physiological features and can reflect the relationship between physiological information. After the overall information is obtained through overall feature extraction, the image is segmented by time and input into the time convolution network, which focuses more on extracting time relationships, and can make physiological detection more accurate.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a physiological information monitoring and early warning method and system. Background Art

[0002] At present, due to the mutual influence of physiological information during detection, the changes in physiological information are different under different external and internal changes. Therefore, when monitoring and detecting it, it is necessary to consider the influence of changes between different physiological information and different time points. However, the current detection of it has not been achieved, resulting in inaccurate detection and judgment of changes in physiological information. Summary of the invention

[0003] The purpose of the present invention is to provide a physiological information monitoring and early warning method and system to solve the above-mentioned problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a physiological information monitoring and early warning method, comprising:

[0005] Acquire multiple physiological information at multiple time points; the multiple physiological information includes heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength;

[0006] Using the time point as the horizontal coordinate and one physiological information as the vertical coordinate, points are drawn on the image to obtain a time physiological image; multiple time physiological images are obtained corresponding to multiple physiological information;

[0007] According to the physiological information, the temporal physiological image is resized to obtain an image of resized size; a plurality of temporal physiological images are correspondingly obtained to a plurality of resized images;

[0008] According to the multiple physiological information, determine the relationship between the physiological information, adjust the positions of the multiple resized images, and obtain an adjusted image;

[0009] The abnormality recognition network predicts abnormal physiological parameters based on the adjusted image input, and issues warning information if the predicted physiological parameter changes abnormally;

[0010] Among them, the abnormal physiological parameters are predicted based on the adjusted image input through the abnormal recognition network, including:

[0011] Input the adjusted image into the first convolutional network to obtain the overall features;

[0012] The adjusted image is divided into time segments and input into the temporal convolutional network to obtain partial features;

[0013] The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological parameters are normal.

[0014] Optionally, the anomaly recognition network includes a first convolutional network, a temporal convolutional network, a pyramid structure, and a classification head;

[0015] The input of the temporal convolutional network is the segmented adjusted image, the input of the first convolutional network is the adjusted image, the input of the pyramid structure is the output of the temporal convolutional network and the output of the first convolutional network; the input of the classification head is the output of the pyramid structure;

[0016] The horizontal coordinates of the adjusted image correspond to different time points, and the adjusted image is evenly divided longitudinally according to the time points to obtain a time segmentation image; the time segmentation image represents the characteristics of multiple physiological information corresponding to a time point reflected in the image;

[0017] Multiple time-segmented images are obtained corresponding to multiple time points;

[0018] Arrange the multiple time-segmented images in order from far to near according to the time points as input of the temporal convolutional network;

[0019] Through the temporal convolutional network, the relationship between various physiological information is extracted from far to near according to the time points, and the relationship between different time points is fused to obtain some features;

[0020] Input the adjusted image into the first convolutional network, extract the overall features of the adjusted image including the time point, the relationship between the physiological information and the degree of influence of the physiological information, and obtain the overall features; the features of the time point are reflected in the horizontal coordinate of the adjusted image, the features of the relationship between the physiological information are reflected in the position of different physiological information in the adjusted image, and the features of the degree of influence of the physiological information are reflected in the size of the corresponding adjusted size image;

[0021] The overall features and partial features are multi-scale fused through a pyramid structure to obtain fused features;

[0022] The fused features are input into a classification head for classification to determine whether the changes in physiological parameters are normal; the output of the classification head is two categories: normal and abnormal.

[0023] Optionally, judging the relationship between the physiological information according to the multiple physiological information, adjusting the positions of the multiple resized images to obtain an adjusted image, includes:

[0024] Obtain similarity values ​​between two physiological information, determine the relationship between two physiological information, and obtain similarity of physiological information; obtain similarity between two physiological information at the same time point for multiple times, and obtain similarity of multiple physiological information; the higher the similarity of the physiological information, the closer the influence of the two physiological information on the abnormal change of physiological parameters is;

[0025] According to the similarity of multiple physiological information, multiple resized images are labeled respectively;

[0026] According to the similarity of the physiological information and in accordance with a set rule, obtaining a position serial number corresponding to the resized image; the position serial number indicates the order of the resized image in the adjusted image;

[0027] Arrange the multiple resized images vertically according to the position sequence numbers from small to large to obtain an adjusted image;

[0028] The setting rules are:

[0029] Arrange the adjusted images with higher similarity values ​​than others in the middle and mark their position numbers;

[0030] Arrange the two adjusted images corresponding to similar values ​​on top of the previously arranged images, using the two in the middle and bottom together;

[0031] Arrange two adjustment images corresponding to similarity values, one above the middle adjustment image and the other below the adjustment image, and mark the position numbers; arrange adjustment images corresponding to multiple similarity values ​​multiple times based on the middle adjustment image, and mark multiple position numbers.

[0032] Optionally, resizing the temporal physiological image according to the physiological information to obtain the resized image includes:

[0033] Obtaining a plurality of physiological value ranges; the physiological value ranges are normal or abnormal values ​​that can be reached by most human bodies corresponding to the physiological information;

[0034] Construct physiological curves based on physiological information at multiple time points;

[0035] Obtaining the variance of the physiological curve to obtain a curve fluctuation value; the curve fluctuation value indicates the degree of fluctuation of the physiological curve;

[0036] According to the physiological value range and the curve fluctuation value, the temporal physiological image is longitudinally adjusted to obtain an adjusted size image.

[0037] Optionally, labeling the multiple resized images respectively according to the similarities of the multiple physiological information includes:

[0038] Obtaining a first similarity; wherein the first similarity is a physiological information similarity that is greater than other similarities among multiple physiological information similarities;

[0039] The two resized images corresponding to the first similarity are marked as 1 and 2 respectively;

[0040] Obtaining a second similarity; wherein the second similarity is a physiological information similarity greater than other similarities among a plurality of physiological information similarities except the physiological information similarity corresponding to the first similarity;

[0041] If a resized image corresponding to the second similarity is the same as a resized image corresponding to the first similarity, marking the other resized image corresponding to the second similarity as 3;

[0042] If the resized image corresponding to the second similarity is different from the resized image corresponding to the first similarity, marking the two resized images corresponding to the second similarity as 3 and 4 respectively;

[0043] Repeat the above method until all resized images are labeled.

[0044] Optionally, the step of arranging the plurality of resized images vertically from small to large according to the position sequence numbers to obtain an adjusted image includes:

[0045] Obtain a first label; the first label is a label greater than other labels; the first label is a label corresponding to the resized image;

[0046] Obtain two position numbers; the two position numbers include a low-order number and a high-order number; the low-order number is 1; the high-order number is the number of resized images;

[0047] Setting the position sequence number of the resized image corresponding to the first label to the average of the high-order sequence number and the low-order sequence number to obtain an intermediate sequence number;

[0048] According to the position sequence number and the number of the resized image, the position sequence number is set based on the intermediate sequence number to obtain resized images of multiple position sequence numbers;

[0049] The resized images are sorted vertically from small to large according to a plurality of position numbers to obtain an adjusted image.

[0050] Optionally, the step of setting the position number based on the position number and the number of the resized image with the intermediate number as a reference to obtain the resized images with multiple position numbers includes:

[0051] Obtaining a second label; the second label is a label greater than other labels except the first label;

[0052] Set the position sequence number of the resized image corresponding to the second label to the low-order sequence number;

[0053] Obtaining a third label; the third label is a label greater than other labels except the first label and the second label;

[0054] The position sequence number of the resized image corresponding to the third label is set to the position sequence number of the middle sequence number plus 1;

[0055] Repeat the above method, adding 1 to the low-order sequence number and the middle sequence number in each loop until all resized images are traversed.

[0056] Optionally, resizing the temporal physiological image according to the physiological value range and the curve fluctuation value to obtain the resized image includes:

[0057] Obtaining a physiological value range ratio and a curve fluctuation value ratio; the physiological value range ratio represents the proportion of the importance of the physiological value range for adjusting the image size; the curve fluctuation value ratio represents the proportion of the importance of the curve fluctuation value for adjusting the image size;

[0058] The physiological value range is multiplied by the physiological value range ratio, and the curve fluctuation value is multiplied by the curve fluctuation value ratio to obtain an adjustment ratio;

[0059] According to the adjustment ratio, the temporal physiological image is vertically stretched by interpolation to obtain an adjusted size image.

[0060] Optionally, the step of plotting points on an image using a time point as a horizontal coordinate and a type of physiological information as a vertical coordinate to obtain a time physiological image includes:

[0061] Obtaining the number of time points; the number of time points is the number of time points at which physiological information is collected;

[0062] Obtaining a horizontal ratio and a vertical ratio of an image; the horizontal ratio of the image represents a ratio of the length of the generated temporal physiological image to the number of time points; the vertical ratio of the image represents a ratio of the width of the generated temporal physiological image to the range of physiological information;

[0063] Multiplying the number of time points by the horizontal ratio of the image to obtain the width of the temporal physiological image;

[0064] The extreme values ​​of the physiological information at multiple time points are multiplied by the longitudinal scale of the image to obtain the length of the temporal physiological image;

[0065] constructing a temporal physiological image according to a temporal physiological image width and a temporal physiological image length;

[0066] The physiological information corresponding to multiple time points is plotted on the time physiological image.

[0067] In a second aspect, an embodiment of the present invention provides a physiological information monitoring and early warning system, including:

[0068] Acquisition module: acquires a variety of physiological information at multiple time points; the various physiological information includes heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength;

[0069] Image construction module: With the time point as the horizontal coordinate and one physiological information as the vertical coordinate, points are drawn on the image to obtain a time physiological image; multiple time physiological images are obtained corresponding to multiple physiological information;

[0070] Image adjustment module: according to the physiological information, resize the time physiological image to obtain an adjusted size image; obtain multiple adjusted size images corresponding to multiple time physiological images; according to the multiple physiological information, determine the relationship between the physiological information, adjust the positions of the multiple adjusted size images, and obtain an adjusted image;

[0071] Image prediction and warning module: Through the abnormal recognition network, the abnormal physiological parameters are predicted based on the adjusted image input. If the predicted physiological parameters change abnormally, a warning message is issued;

[0072] Among them, the abnormal physiological parameters are predicted based on the adjusted image input through the abnormal recognition network, including:

[0073] Input the adjusted image into the first convolutional network to obtain the overall features;

[0074] The adjusted image is divided into time segments and input into the temporal convolutional network to obtain partial features;

[0075] The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological parameters are normal.

[0076] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0077] The embodiment of the present invention also provides a physiological information monitoring and early warning method and system, the method comprising: obtaining multiple physiological information at multiple time points. The multiple physiological information includes heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength. With the time point as the horizontal coordinate and one physiological information as the vertical coordinate, points are drawn on the image to obtain a time physiological image. Multiple time physiological images are obtained corresponding to the multiple physiological information. According to the physiological information, the time physiological image is resized to obtain an adjusted size image. Multiple time physiological images are obtained corresponding to multiple adjusted size images. According to the multiple physiological information, the relationship between the physiological information is judged, and the multiple adjusted size images are adjusted in position to obtain an adjusted image. The abnormal physiological parameter is predicted based on the adjusted image input by the abnormal recognition network, and if the physiological parameter change is predicted to be abnormal, a warning message is issued. Among them, the abnormal physiological parameter is predicted based on the adjusted image input by the abnormal recognition network, including: inputting the adjusted image into the first convolutional network to obtain the overall feature. The adjusted image is divided according to time and input into the time convolutional network to obtain partial features. The overall feature and the partial feature are fused and input into the classification head to judge whether the physiological parameter change is normal.

[0078] In the present invention, the physiological information is converted to an image, and more physiological information features can be added according to the adjustment of the image. According to the longitudinal size change and the position change of multiple resized images, the feature information is transferred to an image, and it is convenient to extract features later. The position sequence is arranged according to the relationship of the physiological information, and the relationship of the physiological information is placed in the middle, and the related one is at the boundary below and the other is above the middle, so that similar physiological information can be in two positions without losing information during feature extraction. Adjusting the size of the time physiological image according to different types of values ​​can make it more accurate to extract physiological features when multiple time physiological images are combined into an adjusted image, and can reflect the relationship between physiological information. After the overall information is obtained through overall feature extraction, the image is segmented by time and input into the time convolution network, which focuses more on extracting time relationships, and can make physiological detection more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of a physiological information monitoring and early warning method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0080] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0081] Example 1

[0082] like Figure 1 As shown, an embodiment of the present invention provides a physiological information monitoring and early warning method, the method comprising:

[0083] S101: Acquire multiple physiological information at multiple time points, including heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength.

[0084] S102: Using the time point as the horizontal coordinate and one physiological information as the vertical coordinate, plot points on the image to obtain a time physiological image. Multiple time physiological images can be obtained corresponding to multiple physiological information.

[0085] A two-dimensional coordinate axis is constructed, with the lower left corner of the time physiological image as the origin, the width direction of the time physiological image as the abscissa, and the length direction of the time physiological image as the ordinate.

[0086] Wherein, one horizontal coordinate of the temporal physiological image corresponds to only one value of physiological information.

[0087] S103: According to the physiological information, the temporal physiological image is resized to obtain a resized image. A plurality of resized images are obtained corresponding to a plurality of temporal physiological images.

[0088] S104: judging the relationship between the physiological information according to the multiple physiological information, and adjusting the positions of the multiple temporal physiological images to obtain an adjusted image.

[0089] S105: Predict physiological parameter abnormalities based on the adjusted image input through an abnormality recognition network, and issue a warning message if the predicted physiological parameter changes abnormally.

[0090] Among them, the abnormal physiological parameters are predicted based on the adjusted image input through the abnormal recognition network, including:

[0091] The adjusted image is input into the first convolutional network to obtain the overall features.

[0092] Among them, in this embodiment, the first convolutional network is a trained convolutional neural network (Convolutional Neural Network, CNN).

[0093] The adjusted image is divided into time segments and input into the temporal convolutional network to obtain some features.

[0094] Among them, a modified temporal convolutional network (TCN) is used in this embodiment.

[0095] The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological parameters are normal.

[0096] Among them, the overall features and partial features are integrated through the pyramid structure.

[0097] Among them, according to deep learning, the accuracy of these physiological information for judgment can be obtained, and this accuracy can be judged through an image containing multiple information.

[0098] Optionally, the anomaly recognition network includes a first convolutional network, a temporal convolutional network, a pyramid structure, and a classification head.

[0099] The input of the temporal convolutional network is the segmented adjusted image, the input of the first convolutional network is the adjusted image, the input of the pyramid structure is the output of the temporal convolutional network and the output of the first convolutional network, and the input of the classification head is the output of the pyramid structure.

[0100] The horizontal coordinates of the adjusted image correspond to different time points, and the adjusted image is evenly divided vertically according to the time points to obtain a time segmentation image. The time segmentation image represents the characteristics of multiple physiological information corresponding to a time point reflected in the image.

[0101] Multiple time-divided images are obtained corresponding to multiple time points.

[0102] The multiple time-segmented images are arranged in sequence from far to near according to the time points as input of the temporal convolutional network.

[0103] The temporal convolutional network is used to extract the relationship between various physiological information from far to near according to the time points, and the relationships at different time points are fused to obtain partial features.

[0104] The adjusted image is input into the first convolutional network, and the overall features including the time point, the relationship between the physiological information and the degree of influence of the physiological information are extracted from the adjusted image to obtain the overall features. The features of the time point are reflected in the horizontal coordinate of the adjusted image, the features of the relationship between the physiological information are reflected in the position of different physiological information in the adjusted image, and the features of the degree of influence of the physiological information are reflected in the size of the corresponding adjusted size image.

[0105] The overall features and partial features are fused at multiple scales through a pyramid structure to obtain fused features.

[0106] The fused features are input into the classification head for classification to determine whether the changes in physiological parameters are normal. The output of the classification head is two categories: normal and abnormal.

[0107] Optionally, judging the relationship between the physiological information according to the multiple physiological information, adjusting the positions of the multiple resized images to obtain an adjusted image, includes:

[0108] Obtain similarity values ​​between two physiological information, determine the relationship between two physiological information, and obtain physiological information similarity. Obtain similarities between two physiological information at the same time point multiple times to obtain multiple physiological information similarities. The higher the physiological information similarity, the closer the impact of the two physiological information on the abnormal change of physiological parameters is.

[0109] According to the similarity of multiple physiological information, multiple resized images are labeled respectively.

[0110] According to the physiological information similarity and the set rule, the position sequence number corresponding to the resized image is obtained. The position sequence number indicates the order of the resized image in the adjusted image.

[0111] Arrange the plurality of resized images vertically according to the position sequence numbers from small to large to obtain an adjusted image.

[0112] The setting rules are:

[0113] Arrange the adjusted images with higher similarity values ​​than others in the middle and mark their position numbers.

[0114] Arrange the two adjusted images corresponding to similar values ​​together on top of the previously arranged images, using the two in the middle and bottom.

[0115] Arrange the two adjustment images corresponding to the similarity values, one above the middle adjustment image and the other below the adjustment image, and mark the position numbers. Arrange the adjustment images corresponding to multiple similarity values ​​multiple times based on the middle adjustment image, and mark multiple position numbers.

[0116] According to the labels of the resized images, the longitudinal positions of the plurality of resized images are adjusted based on the relationship of physiological information to obtain an adjusted image.

[0117] Through the above method, the longitudinal size change and the position change of the multiple resized images are performed according to the multiple resized images, and the feature information is transferred to one image, which is convenient for subsequent feature extraction.

[0118] Optionally, labeling the multiple resized images respectively according to the similarities of the multiple physiological information includes:

[0119] The similarity of two physiological information at the same time point is obtained to obtain the physiological information similarity. The similarity between two of the multiple physiological information at the same time point is obtained multiple times to obtain multiple physiological information similarities.

[0120] Among them, the similarity is calculated by Euclidean distance.

[0121] In this embodiment, the number of resized images is 5, and the number of physiological information similarities is 5*4 / 2=10.

[0122] The two resized images corresponding to the first similarity are marked as 1 and 2 respectively.

[0123] A second similarity is obtained, wherein the second similarity is a physiological information similarity greater than other similarities among a plurality of physiological information similarities except the physiological information similarity corresponding to the first similarity.

[0124] If a resized image corresponding to the second similarity is the same as a resized image corresponding to the first similarity, the other resized image corresponding to the second similarity is marked as 3.

[0125] Since the similarity between two images is calculated, the resized image corresponding to the second similarity may be the same as the resized image corresponding to the first similarity, so a judgment is required.

[0126] If the resized image corresponding to the second similarity is different from the resized image corresponding to the first similarity, the two resized images corresponding to the second similarity are marked as 3 and 4 respectively.

[0127] Repeat the above method until all resized images are labeled.

[0128] In this embodiment, the number of resized images is 5, and the third similarity is obtained again. It is determined whether the resized image corresponding to the third similarity is in the resized images corresponding to the first similarity and the second similarity.

[0129] If yes, mark the other resized image corresponding to the second similarity with the number at the end of the first cycle plus 1. If the resized image corresponding to the second similarity is the same as the resized image corresponding to the first similarity, mark the other resized image corresponding to the second similarity as 3, and mark the different resized images starting from 4.

[0130] Optionally, the step of arranging the plurality of resized images vertically from small to large according to the position sequence numbers to obtain an adjusted image includes:

[0131] Obtain a first label. The first label is a label greater than other labels. The first label is the label corresponding to the resized image

[0132] Two position numbers are obtained. The two position numbers include a low-order number and a high-order number. The low-order number is 1. The high-order number is the number of resized images.

[0133] The low-order serial number indicates the position below the adjusted image, and the high-order serial number indicates the position above the adjusted image.

[0134] The position serial number of the resized image corresponding to the first label is set to the average of the high-order serial number and the low-order serial number to obtain an intermediate serial number.

[0135] When calculating the average value, the value is rounded down.

[0136] In this embodiment, if the number of resized images is 5, the first label is 5. The position number of the resized image labeled 5 is set to 3.

[0137] According to the position serial number and the number of the resized image, the position serial number is set based on the intermediate serial number to obtain resized images of multiple position serial numbers.

[0138] The resized images are sorted vertically from small to large according to a plurality of position numbers to obtain an adjusted image.

[0139] Through the above method, the position numbers are arranged according to the relationship of physiological information, the physiological information is placed in the middle, the related one is at the boundary below and the other is above the middle, so that similar physiological information can be in two positions without losing information during feature extraction. Adjusting the size of the temporal physiological image according to different types of values ​​can make it more accurate to extract physiological features when combining multiple temporal physiological images into one adjusted image, and can reflect the relationship between physiological information.

[0140] Optionally, the step of setting the position number based on the position number and the number of the resized image with the intermediate number as a reference to obtain the resized images with multiple position numbers includes:

[0141] A second label is obtained, wherein the second label is a label greater than other labels except the first label.

[0142] The first label is the label corresponding to the resized image.

[0143] In this embodiment, if the number of resized images is 5, then the second label is 4.

[0144] The position sequence number of the resized image corresponding to the second label is set to a low-order sequence number.

[0145] The position number of the resized image labeled 4 is set to 1.

[0146] A third label is obtained, wherein the third label is a label greater than other labels except the first label and the second label.

[0147] In this embodiment, if the number of resized images is 5, then the third label is 3.

[0148] The position serial number of the resized image corresponding to the third label is set to the position serial number of the middle serial number plus 1.

[0149] The position number of the resized image labeled 3 is set to 4.

[0150] Repeat the above method, adding 1 to the low-order sequence number and the middle sequence number in each loop until all resized images are traversed.

[0151] In this embodiment, the number of resized images is 5, so the position numbers of other images are:

[0152] The position number of the resized image corresponding to the fourth number is set to the position number of the low number plus 1. The position number of the resized image corresponding to the fifth number is set to the position number of the middle number plus 2. The position number of the resized image with the number 2 is set to 2. The position number of the resized image with the number 1 is set to 5.

[0153] Through the above method, the changes in physiological parameters (some changes in physiological parameters are related and some are not related) are obtained.

[0154] Optionally, resizing the temporal physiological image according to the physiological information to obtain the resized image includes:

[0155] A plurality of physiological value ranges are obtained, wherein the physiological value ranges are normal or abnormal values ​​that can be reached by most human bodies corresponding to the physiological information.

[0156] Construct a physiological curve based on physiological information at multiple time points.

[0157] In this embodiment, the fitting is performed by polynomial fitting.

[0158] The variance of the physiological curve is obtained to obtain a curve fluctuation value, wherein the curve fluctuation value indicates the degree of fluctuation of the physiological curve.

[0159] According to the physiological value range and the curve fluctuation value, the temporal physiological image is resized to obtain an image of resized size.

[0160] Optionally, resizing the temporal physiological image according to the physiological value range, the curve fluctuation value and the sensitivity value to obtain the resized image includes:

[0161] A physiological value range ratio and a curve fluctuation value ratio are obtained. The physiological value range ratio indicates the proportion of the importance of the physiological value range for adjusting the image size. The curve fluctuation value ratio indicates the proportion of the importance of the curve fluctuation value for adjusting the image size.

[0162] The adjustment ratio is obtained by multiplying the physiological value range by the physiological value range ratio and the curve fluctuation value by the curve fluctuation value ratio.

[0163] According to the adjustment ratio, the temporal physiological image is vertically stretched by interpolation to obtain an adjusted size image.

[0164] Among them, the time physiological image is vertically stretched through the resize function in the Image algorithm of the PIL library.

[0165] Optionally, the step of plotting points on an image using a time point as a horizontal coordinate and a type of physiological information as a vertical coordinate to obtain a time physiological image includes:

[0166] Obtain the number of time points, which is the number of time points at which physiological information is collected.

[0167] The image horizontal ratio and the image vertical ratio are obtained. The image horizontal ratio represents the ratio of the length of the generated temporal physiological image to the number of time points. The image vertical ratio represents the ratio of the width of the generated temporal physiological image to the range of physiological information.

[0168] The width of the temporal physiological image is obtained by multiplying the number of time points by the horizontal scale of the image.

[0169] The extreme values ​​of the physiological information at multiple time points are multiplied by the longitudinal scale of the image to obtain the length of the temporal physiological image.

[0170] A temporal physiological image is constructed according to a temporal physiological image width and a temporal physiological image length.

[0171] The physiological information corresponding to multiple time points is plotted on the time physiological image.

[0172] The step of multiplying the extreme values ​​of the physiological information at multiple time points by the longitudinal ratio of the image to obtain the length of the temporal physiological image includes:

[0173] A high physiological value is obtained, wherein the high physiological value is a value of physiological information greater than other values ​​at multiple time points.

[0174] A low physiological value is obtained. The high physiological value is a value of physiological information that is smaller than other values ​​at multiple time points.

[0175] Subtract the low physiological value from the high physiological value to obtain the length of the temporal physiological image.

[0176] Example 2

[0177] Based on the above-mentioned physiological information monitoring and early warning method, an embodiment of the present invention further provides a physiological information monitoring and early warning system, which includes an acquisition module, an image construction module, an image adjustment module and an image prediction and early warning module.

[0178] The acquisition module is used to acquire multiple physiological information at multiple time points, including heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength.

[0179] The image construction module is used to draw points on the image with the time point as the horizontal coordinate and one physiological information as the vertical coordinate to obtain a time physiological image. Multiple time physiological images are obtained corresponding to multiple physiological information.

[0180] The image adjustment module is used to adjust the size of the time physiological image according to the physiological information to obtain an adjusted size image. A plurality of time physiological images correspond to a plurality of adjusted size images. According to the plurality of physiological information, the relationship of the physiological information is determined, and the plurality of adjusted size images are adjusted in position to obtain an adjusted image.

[0181] The image prediction and warning module is used to predict abnormal physiological parameters based on the adjusted image input through the abnormal recognition network. If the predicted physiological parameters change abnormally, a warning message is issued.

[0182] Among them, the abnormal physiological parameters are predicted based on the adjusted image input through the abnormal recognition network, including:

[0183] The adjusted image is input into the first convolutional network to obtain the overall features.

[0184] The adjusted image is divided into time segments and input into the temporal convolutional network to obtain some features.

[0185] The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological parameters are normal.

[0186] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0187] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0188] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0189] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0190] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0191] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in the apparatus according to an embodiment of the present invention. The present invention may also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may have the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0192] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc., does not indicate any order. These words may be interpreted as names.

Claims

1. A physiological information monitoring and early warning method, characterized in that: include: Obtain a variety of physiological information at multiple time points; The multiple physiological information includes heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength; Using the time point as the horizontal coordinate and a physiological information as the vertical coordinate, points are drawn on the image to obtain a time physiological image; Multiple physiological information corresponds to multiple temporal physiological images; According to a kind of physiological information, the temporal physiological image is resized to obtain an resized image; Multiple time physiological images correspond to multiple resized images; According to the multiple physiological information, determine the relationship between the multiple physiological information, adjust the positions of the multiple resized images, and obtain an adjusted image; The adjusted image is input into the abnormality recognition network to predict abnormal physiological information. If the predicted physiological information changes abnormally, a warning message is issued; Among them, the adjusted image is input into the abnormality recognition network to predict abnormal physiological information, including: Input the adjusted image into the first convolutional network to obtain the overall features; The adjusted image is divided into time segments and input into the temporal convolutional network to obtain partial features; The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological information are normal; The anomaly recognition network includes the first convolutional network, the temporal convolutional network, the pyramid structure and the classification head; The input of the temporal convolutional network is the segmented adjusted image, the input of the first convolutional network is the adjusted image, the input of the pyramid structure is the output of the temporal convolutional network and the output of the first convolutional network; the input of the classification head is the output of the pyramid structure; The horizontal coordinates of the adjusted image correspond to different time points, and the adjusted image is evenly divided longitudinally according to the time points to obtain a time segmentation image; the time segmentation image represents the characteristics of multiple physiological information corresponding to a time point reflected in the image; Multiple time-segmented images are obtained corresponding to multiple time points; Arrange the multiple time-segmented images in order from far to near according to the time points as input of the temporal convolutional network; Through the temporal convolutional network, the relationship between multiple physiological information is extracted from far to near according to the time points, and the relationship between multiple physiological information at different time points is fused to obtain partial features; Input the adjusted image into the first convolutional network, extract the overall features of the adjusted image including the time point, the relationship between the physiological information and the degree of influence of the physiological information, and obtain the overall features; the features of the time point are reflected in the horizontal coordinate of the adjusted image, the features of the relationship between the physiological information are reflected in the position of different physiological information in the adjusted image, and the features of the degree of influence of the physiological information are reflected in the size of the corresponding adjusted size image; The overall features and partial features are multi-scale fused through a pyramid structure to obtain fused features; The fused features are input into a classification head for classification to determine whether the physiological information changes are normal; the output of the classification head is two classifications of normal and abnormal; The method of determining the relationship between the physiological information according to the multiple physiological information and adjusting the positions of the multiple resized images to obtain an adjusted image includes: Obtain similarity values ​​between two physiological information, determine the relationship between two physiological information, and obtain similarity of physiological information; obtain similarity between two physiological information at the same time point for multiple times, and obtain similarity of multiple physiological information; the higher the similarity of physiological information, the closer the influence of two physiological information on abnormal changes of physiological information; According to the similarity of multiple physiological information, multiple resized images are labeled respectively; According to the similarities of the multiple physiological information and the set rules, obtaining the position serial number corresponding to the resized image; the position serial number indicates the order of the resized image in the adjusted image; Arrange the multiple resized images vertically according to the position sequence numbers from small to large to obtain an adjusted image; The setting rules are: Arrange the resized images with higher similarity values ​​than others in the middle and mark their position numbers; Arrange one of the two resized images corresponding to the similarity values ​​above the middle resized image, and arrange the other resized image below the middle resized image, and mark the position numbers; arrange the resized images corresponding to multiple similarity values ​​multiple times based on the middle resized image, and mark multiple position numbers; The step of resizing the temporal physiological image according to the physiological information to obtain the resized image includes: Obtaining a plurality of physiological value ranges; the plurality of physiological value ranges are normal or abnormal values ​​corresponding to a plurality of physiological information; Construct physiological curves based on physiological information at multiple time points; Obtaining the variance of the physiological curve to obtain a curve fluctuation value; the curve fluctuation value indicates the degree of fluctuation of the physiological curve; resizing the temporal physiological image according to the plurality of physiological value ranges and curve fluctuation values ​​to obtain an adjusted-size image; The method of vertically arranging the plurality of resized images from small to large according to the position sequence numbers to obtain an adjusted image comprises: Obtaining a first label; the first label is a label greater than other labels; Obtain two position numbers; the two position numbers include a low-order number and a high-order number; the low-order number is 1; the high-order number is the number of resized images; Setting the position sequence number of the resized image corresponding to the first label to the average of the high sequence number and the low sequence number to obtain an intermediate sequence number; According to the position sequence number and the number of the resized image, the position sequence number is set based on the intermediate sequence number to obtain resized images of multiple position sequence numbers; The resized images are sorted vertically from small to large according to a plurality of position numbers to obtain an adjusted image.

2. A physiological information monitoring and early warning method according to claim 1, characterized in that: The step of labeling the plurality of resized images respectively according to the similarities of the plurality of physiological information comprises: Obtaining a first similarity; wherein the first similarity is a physiological information similarity that is greater than other similarities among multiple physiological information similarities; The two resized images corresponding to the first similarity are marked as 1 and 2 respectively; Obtaining a second similarity; wherein the second similarity is a physiological information similarity greater than other similarities among a plurality of physiological information similarities except the physiological information similarity corresponding to the first similarity; If a resized image corresponding to the second similarity is the same as a resized image corresponding to the first similarity, marking the other resized image corresponding to the second similarity as 3; If the resized image corresponding to the second similarity is different from the resized image corresponding to the first similarity, marking the two resized images corresponding to the second similarity as 3 and 4 respectively; Repeat the above steps until all resized images are labeled.

3. A physiological information monitoring and early warning method according to claim 1, characterized in that: The step of setting the position number based on the position number and the number of the resized image with the intermediate number as a reference to obtain the resized images of multiple position numbers includes: Obtaining a second label; the second label is a label greater than other labels except the first label; Set the position sequence number of the resized image corresponding to the second label to the low-order sequence number; Obtaining a third label; the third label is a label greater than other labels except the first label and the second label; The position sequence number of the resized image corresponding to the third label is set to the position sequence number of the middle sequence number plus 1; Repeat the above steps, adding 1 to the low-order sequence number and the middle sequence number each time, until all resized images are traversed.

4. A physiological information monitoring and early warning method according to claim 1, characterized in that: The method of plotting points on an image using a time point as a horizontal coordinate and a type of physiological information as a vertical coordinate to obtain a time physiological image includes: Obtaining the number of time points; the number of time points is the number of time points at which physiological information is collected; Obtaining an image horizontal ratio and an image vertical ratio; the image horizontal ratio represents a ratio of the length of the generated temporal physiological image to the number of time points; the image vertical ratio represents a ratio of the width of the generated temporal physiological image to the range of physiological information; Multiplying the number of time points by the horizontal scale of the image to obtain the length of the temporal physiological image; The width of the temporal physiological image is obtained by multiplying the range of the physiological information at multiple time points by the longitudinal scale of the image; A temporal physiological image is constructed according to a temporal physiological image width and a temporal physiological image length.

5. A physiological information monitoring and early warning system, characterized in that: include: Acquisition module: obtains various physiological information at multiple time points; The multiple physiological information includes heart rate, body temperature, blood oxygen, blood pressure, muscle tension and muscle strength; Image construction module: using the time point as the horizontal coordinate and a physiological information as the vertical coordinate, points are drawn on the image to obtain a time physiological image; Multiple physiological information corresponds to multiple temporal physiological images; An image adjustment module: according to a physiological information, the temporal physiological image is resized to obtain an image of resized image; Obtaining a plurality of resized images corresponding to the plurality of time physiological images; judging the relationship between the plurality of physiological information according to the plurality of physiological information, and adjusting the positions of the plurality of resized images to obtain an adjusted image; Image prediction and warning module: input the adjusted image into the abnormality recognition network to predict abnormal physiological information. If the predicted physiological information changes abnormally, a warning message is issued; Among them, the adjusted image is input into the abnormality recognition network to predict abnormal physiological information, including: Input the adjusted image into the first convolutional network to obtain the overall features; The adjusted image is divided into time segments and input into the temporal convolutional network to obtain partial features; The overall features and partial features are fused and input into the classification head to determine whether the changes in physiological information are normal; The anomaly recognition network includes the first convolutional network, the temporal convolutional network, the pyramid structure and the classification head; The input of the temporal convolutional network is the segmented adjusted image, the input of the first convolutional network is the adjusted image, the input of the pyramid structure is the output of the temporal convolutional network and the output of the first convolutional network; the input of the classification head is the output of the pyramid structure; The horizontal coordinates of the adjusted image correspond to different time points, and the adjusted image is evenly divided longitudinally according to the time points to obtain a time segmentation image; the time segmentation image represents the characteristics of multiple physiological information corresponding to a time point reflected in the image; Multiple time-segmented images are obtained corresponding to multiple time points; Arrange the multiple time-segmented images in order from far to near according to the time points as input of the temporal convolutional network; Through the temporal convolutional network, the relationship between multiple physiological information is extracted from far to near according to the time points, and the relationship between multiple physiological information at different time points is fused to obtain partial features; Input the adjusted image into the first convolutional network, extract the overall features of the adjusted image including the time point, the relationship between the physiological information and the degree of influence of the physiological information, and obtain the overall features; the features of the time point are reflected in the horizontal coordinate of the adjusted image, the features of the relationship between the physiological information are reflected in the position of different physiological information in the adjusted image, and the features of the degree of influence of the physiological information are reflected in the size of the corresponding adjusted size image; The overall features and partial features are multi-scale fused through a pyramid structure to obtain fused features; The fused features are input into a classification head for classification to determine whether the physiological information changes are normal; the output of the classification head is two classifications of normal and abnormal; The method of determining the relationship between the physiological information according to the multiple physiological information and adjusting the positions of the multiple resized images to obtain an adjusted image includes: Obtain similarity values ​​between two physiological information, determine the relationship between two physiological information, and obtain similarity of physiological information; obtain similarity between two physiological information at the same time point for multiple times, and obtain similarity of multiple physiological information; the higher the similarity of physiological information, the closer the influence of two physiological information on abnormal changes of physiological information; According to the similarity of multiple physiological information, multiple resized images are labeled respectively; According to the similarities of the multiple physiological information and the set rules, obtaining the position serial number corresponding to the resized image; the position serial number indicates the order of the resized image in the adjusted image; Arrange the multiple resized images vertically according to the position sequence numbers from small to large to obtain an adjusted image; The setting rules are: Arrange the resized images with higher similarity values ​​than others in the middle and mark their position numbers; Arrange one of the two resized images corresponding to the similarity values ​​above the middle resized image, and arrange the other resized image below the middle resized image, and mark the position numbers; arrange the resized images corresponding to multiple similarity values ​​multiple times based on the middle resized image, and mark multiple position numbers; The step of resizing the temporal physiological image according to the physiological information to obtain the resized image includes: Obtaining a plurality of physiological value ranges; the plurality of physiological value ranges are normal or abnormal values ​​corresponding to a plurality of physiological information; Construct physiological curves based on physiological information at multiple time points; Obtaining the variance of the physiological curve to obtain a curve fluctuation value; the curve fluctuation value indicates the degree of fluctuation of the physiological curve; resizing the temporal physiological image according to the plurality of physiological value ranges and curve fluctuation values ​​to obtain an adjusted-size image; The method of vertically arranging the plurality of resized images from small to large according to the position sequence numbers to obtain an adjusted image comprises: Obtaining a first label; the first label is a label greater than other labels; Obtain two position numbers; the two position numbers include a low-order number and a high-order number; the low-order number is 1; the high-order number is the number of resized images; Setting the position sequence number of the resized image corresponding to the first label to the average of the high sequence number and the low sequence number to obtain an intermediate sequence number; According to the position sequence number and the number of the resized image, the position sequence number is set based on the intermediate sequence number to obtain resized images of multiple position sequence numbers; The resized images are sorted vertically from small to large according to a plurality of position numbers to obtain an adjusted image.

Images