Internet-based breast cancer patient rehabilitation nursing system and method

Abstract

This invention relates to the field of intelligent medical and rehabilitation nursing technology, and discloses an internet-based rehabilitation nursing system and method for breast cancer patients. The system includes: a vital sign modeling module for acquiring resting upper limb images, calculating the edema index, and extracting colorimetric baseline data; a feature extraction module for acquiring motion image sequences to construct spatial skeletal vectors and generating skin tone tension mutation data in conjunction with the baseline data; a compensation identification module for generating a dynamic tolerance threshold based on the edema index, performing dual-gated comparison of the skeletal vectors and tension mutation data, and generating a compensation diagnosis signal; and a prescription control module for issuing a downgraded prescription based on the diagnosis signal to interrupt training, and acquiring probe-state skin tone tension mutation data after intervention, and performing adaptive prescription flow control based on the comparison results. This invention accurately identifies compensatory movements caused by scar tension through multimodal feature cross-validation, effectively preventing secondary injuries caused by home rehabilitation for patients.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical and rehabilitation nursing technology, specifically to an Internet-based rehabilitation nursing system and method for breast cancer patients. Background Technology

[0002] Breast cancer is a malignant tumor that occurs in the glandular epithelial tissue of the breast and is one of the most common malignant tumors in clinical practice. Surgical treatment typically includes resection of the breast lesion and axillary lymph node dissection. Because the surgical area involves a complex network of blood vessels, nerves, and lymphatic vessels, patients are prone to developing subcutaneous fascial adhesions and surgical scars in the affected axilla during postoperative healing, often presenting clinically as axillary-reticulum syndrome. These physiological structural changes can lead to severe limitation of joint mobility and pulling pain in the affected upper limb. If not intervened in time, this can develop into permanent upper limb dysfunction. Therefore, patients must undergo specific upper limb rehabilitation exercises (such as abduction and elevation movements) as early and for a long period after surgery to physically stretch the adhered subcutaneous tissue, reshape muscle elasticity, and promote lymphatic drainage in the affected limb, thereby restoring normal physiological function of the limb.

[0003] With the development of internet-based healthcare, existing home-based rehabilitation care typically relies on general rehabilitation software on smart terminals. This technology primarily uses a monocular camera to capture the patient's two-dimensional skeletal points in real time, assessing whether the patient's limb elevation has reached the set training goals. However, in practice, when a patient's affected limb has limited elevation, in an attempt to forcefully reach the system's set wrist height threshold, the patient will instinctively use incorrect postures such as shrugging their shoulders and laterally flexing their trunk to the healthy side. Existing general visual recognition technology only uses distal limb displacement as the criterion for achieving the target, making it difficult to distinguish whether the patient's inadequate movement is due to simple muscle weakness or laziness, or the presence of a real risk of scar tension. This technological blind spot leads the system to frequently misjudge severe, implicit compensatory behaviors as acceptable movements. Patients not only fail to achieve the rehabilitation goal of stretching limited scars, but may also develop severe cervical and shoulder syndrome and secondary tears in the subcutaneous fascia due to maintaining this incorrect compensatory force pattern for a long time. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an internet-based rehabilitation and nursing system and method for breast cancer patients. This system solves the technical problem that existing technologies struggle to identify the implicit compensatory movements caused by scar traction, which can easily lead to long-term incorrect exertion and secondary injury.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an Internet-based breast cancer patient rehabilitation and nursing system, comprising the following modules:

[0006] The vital signs modeling module is used to acquire images of the patient's upper limbs in a resting state collected by a smart terminal, extract the length of the forearm bones and the pixel area of ​​the forearm, normalize and calculate the edema index, and perform color space conversion on the region of interest in the upper limb image to extract chromaticity reference data.

[0007] The feature extraction module is used to acquire motion image sequences of patients performing upper limb rehabilitation movements in real time collected by the smart terminal, construct spatial skeleton vectors through the coordinates of key human nodes, extract the current chromaticity data of the region of interest in the motion image sequence, and generate skin tone tension mutation data by comparing with the chromaticity reference data.

[0008] The compensation identification module is used to calculate the scapular compensation angle based on the spatial skeletal vector, generate a dynamic tolerance threshold based on the edema index, and perform a dual-gated comparison with the skin tension mutation data and the scapular compensation angle to generate a compensation diagnosis signal.

[0009] The prescription control module is used to send an interruption command to the smart terminal to stop the current rehabilitation training after receiving the compensatory diagnosis signal, issue a downgraded rehabilitation prescription, and issue a detection command to obtain the detection state skin tension mutation data after the downgrade intervention cycle ends, and perform adaptive flow control of the rehabilitation prescription based on the data comparison results;

[0010] By adopting the above technical solution, the system first acquires images of the patient's upper limbs at rest, calculates the edema index by extracting the forearm bone length and forearm pixel area and normalizing them, and extracts colorimetric reference data. When the patient performs rehabilitation movements, the system acquires motion image sequences to construct spatial bone vectors and generates skin tone tension mutation data by comparing them with the colorimetric reference data. As a result, the system can generate a dynamic tolerance threshold based on the edema index and perform dual gating comparisons by combining the skin tone tension mutation data with the scapular compensation angle calculated based on the spatial bone vectors, thereby accurately generating a compensation diagnosis signal. Based on this signal, the system can promptly issue an interruption command to stop the current rehabilitation training and issue a downgraded rehabilitation prescription. After the downgraded intervention cycle ends, the system acquires the probe-state skin tone tension mutation data and finally effectively executes the adaptive flow control of the rehabilitation prescription based on the data comparison results.

[0011] This invention provides an internet-based rehabilitation and nursing system and method for breast cancer patients. It has the following beneficial effects:

[0012] 1. This invention solves the problem that existing general rehabilitation software cannot recognize the real-world situation where patients instinctively take implicit compensatory actions such as shrugging and lateral flexion of the trunk due to axillary scar traction pain by performing a dual-gated comparison of spatial skeletal vectors reflecting postural displacement and skin tension mutation data reflecting subcutaneous tissue physical stretching. By using cross-validation of skeletal kinematic characteristics and local capillary compression whitening characteristics, the system can distinguish between simple muscle weakness and laziness and real dangerous scar stretching, thus solving the risk of secondary injuries such as cervical and shoulder syndrome caused by long-term incorrect force compensation.

[0013] 2. This invention calculates the edema index by acquiring images of the patient's upper limbs at rest and automatically generates a dynamic tolerance threshold based on this. This solves the problem of severe psychological frustration experienced by breast cancer patients who cannot complete fixed rehabilitation exercise goals due to varying degrees of lymphedema each day. The system can adaptively relax the judgment boundary for qualified movements based on the patient's limb swelling and joint restriction on the day, effectively maintaining the patient's enthusiasm for persisting in rehabilitation training at home while avoiding muscle strain caused by forced stretching.

[0014] 3. This invention addresses the problem of patients not knowing how to safely handle sudden traction pain when they encounter it at home without medical care by immediately issuing a downgraded rehabilitation prescription to interrupt training after the diagnosis of compensation, and by performing adaptive flow control after the downgraded intervention by comparing the skin tension mutation data of the detection state. It not only uses low-intensity intervention to promptly block high-risk stretching, but also uses objective tension retest data to determine whether to resume advanced training or forcibly terminate the warning, thus improving the clinical reliability of the Internet remote rehabilitation system. Attached Figure Description

[0015] Figure 1 This is an architecture diagram of the Internet-based breast cancer patient rehabilitation and nursing system of the present invention;

[0016] Figure 2 This is an architecture diagram of the vital sign modeling module provided in an embodiment of the present invention;

[0017] Figure 3 This is an internal architecture diagram of the feature extraction module in an embodiment of the present invention;

[0018] Figure 4 This is a flowchart of the Internet-based rehabilitation and nursing method for breast cancer patients according to the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] Example 1

[0021] Please see the appendix Figure 1 - Appendix Figure 3 This invention provides an internet-based rehabilitation and nursing system for breast cancer patients, comprising the following modules:

[0022] The vital signs modeling module is used to acquire images of the patient's upper limbs in a resting state collected by a smart terminal, extract the length of the forearm bones and the pixel area of ​​the forearm, normalize and calculate the edema index, and perform color space conversion on the region of interest in the upper limb image to extract chromaticity reference data.

[0023] Furthermore, the vital sign modeling module includes:

[0024] The image acquisition unit is used to acquire images of both upper limbs in a resting state, extract the coordinates of key skeletal points using a two-dimensional human pose estimation model, and locate the region of interest in the image of the affected upper limb based on the coordinates of the key skeletal points.

[0025] The edema calculation unit is used to extract the pixel length of the bones of both forearms based on the coordinates of the skeletal key points, crop the image regions of both forearms based on the coordinates of the skeletal key points, and perform normalization calculation by combining the pixel area of ​​the image regions of both forearms to obtain the edema index.

[0026] The reference extraction unit is used to map the image data of the region of interest from the RGB color space to the YCbCr color space, strip the luminance channel data, and retain the chrominance channel data as chrominance reference data.

[0027] The edema calculation unit is specifically used for:

[0028] Extract the pixel length of the bones in the affected forearm and the pixel length of the bones in the healthy forearm;

[0029] Extract the pixel area of ​​the affected forearm image region and the pixel area of ​​the healthy forearm image region from the bilateral forearm image regions;

[0030] Divide the pixel area of ​​the affected forearm image region by the square of the pixel length of the affected forearm bone to obtain the normalized arm circumference feature value of the affected side.

[0031] Divide the pixel area of ​​the healthy forearm image region by the square of the pixel length of the healthy forearm bone to obtain the normalized arm circumference feature value of the healthy side.

[0032] Calculate the difference between the normalized arm circumference characteristic value of the affected side and the normalized arm circumference characteristic value of the healthy side, and divide the difference by the normalized arm circumference characteristic value of the healthy side to obtain the edema index;

[0033] The reference extraction unit is specifically used for:

[0034] Map the image data of the region of interest from the RGB color space to the YCbCr color space;

[0035] Remove the Y channel data from the YCbCr color space;

[0036] The Cb channel data is extracted to construct the blue hue difference benchmark matrix, and the Cr channel data is extracted to construct the red hue difference benchmark matrix.

[0037] The chromaticity reference data is obtained by combining the blue hue difference reference matrix and the red hue difference reference matrix;

[0038] Specifically, after receiving a frontal view image of both upper limbs in a resting state, the image acquisition unit inputs the pixel matrix of the image into a pre-configured human two-dimensional pose estimation model. This model extracts the image edge and texture feature matrix through an internal convolutional operation network, regresses and calculates the horizontal and vertical coordinates of the pixels of each major joint of the human body in the two-dimensional image coordinate system, and outputs the skeletal key point coordinate sequence. The extracted skeletal key point coordinates include acromion node coordinates, elbow joint node coordinates, and wrist joint node coordinates. Based on the linear geometric relationship and spatial distribution of the acromion node coordinates and elbow joint node coordinates, the image acquisition unit extracts a specific range of pixel blocks in the two-dimensional pixel array of the affected upper limb image, determines the pixel block as the region of interest, and transmits the pixel coordinate boundary values ​​of the region of interest to the reference extraction unit.

[0039] The edema calculation unit receives the coordinates of the aforementioned skeletal key points and the image data of both upper limbs. When extracting the pixel length of the forearm bones, the edema calculation unit extracts the x and y coordinates of the elbow joint node on the affected side, as well as the x and y coordinates of the wrist joint node on the affected side. Based on the principle of distance between two points in a two-dimensional plane, the pixel length of the forearm bones on the affected side is calculated using the following formula:

[0040] ;

[0041] in, The pixel length representing the bones of the affected forearm. The x-coordinate represents the elbow joint node on the affected side. The x-coordinate represents the node of the affected wrist joint. The vertical coordinate represents the elbow joint node on the affected side. The vertical coordinate of the wrist joint node represents the ordinate of the affected side. Similarly, the edema calculation unit extracts the pixel length of the healthy forearm bone based on the square root of the sum of the squares of the differences between the horizontal and vertical coordinates of the healthy elbow joint node and the healthy wrist joint node. Then, based on the closed area enclosed by the line connecting the elbow joint node coordinate and the wrist joint node coordinate, the edema calculation unit extracts the image region matrix of the affected forearm and the image region matrix of the healthy forearm from the bilateral upper limb images, respectively, using a pixel segmentation algorithm. The edema calculation unit extracts the pixel area of ​​the image region of the affected forearm and the pixel area of ​​the image region of the healthy forearm by accumulating and counting the total number of pixels in these two independent image matrices.

[0042] The edema calculation unit divides the pixel area of ​​the extracted affected forearm image region by the square of the pixel length of the affected forearm bone to generate the normalized arm circumference feature value for the affected side; simultaneously, it divides the pixel area of ​​the healthy forearm image region by the square of the pixel length of the healthy forearm bone to generate the normalized arm circumference feature value for the healthy side, as shown in the following formula:

[0043] ;

[0044] ;

[0045] in, Represents the normalized arm circumference characteristic value of the affected side. The pixel area representing the image region of the affected forearm. The pixel length representing the bones of the affected forearm. Represents the normalized arm circumference eigenvalue of the healthy side. The pixel area representing the image region of the healthy forearm. The pixel length representing the bone of the healthy forearm;

[0046] The edema calculation unit calculates the difference between the normalized arm circumference characteristic value of the affected side and the normalized arm circumference characteristic value of the healthy side, and then divides this difference by the normalized arm circumference characteristic value of the healthy side to obtain the edema index, as shown in the following formula:

[0047] ;

[0048] in, The calculated edema index is stored in the system register and participates in the subsequent calculation of the dynamic tolerance threshold. The baseline extraction unit acquires the image data of the region of interest. This image data is initially presented as a multi-dimensional RGB color space matrix containing red, green, and blue channel data. Based on a linear color transformation mapping rule, the baseline extraction unit substitutes the channel values ​​of each pixel within the region of interest into the transformation equation, mapping the image data from the RGB color space to the YCbCr color space. The formula is as follows:

[0049] ;

[0050] ;

[0051] ;

[0052] in, Represents the luminance channel data. Represents the blue hue differential channel data. Represents the red-color differential channel data. Represents red channel data, Represents green channel data, Representing the blue channel data, after mapping, the reference extraction unit performs data separation processing on the three-dimensional matrix of the YCbCr color space. The reference extraction unit removes the data derived from the first calculation above. The channel data matrix is ​​used to extract the brightness and darkness attributes from the pixel array, and the baseline extraction unit extracts and retains the relevant parameters. The two-dimensional numerical matrix composed of channel data serves as the blue intensity difference reference matrix, and is extracted synchronously. The two-dimensional numerical matrix composed of channel data serves as the red hue difference reference matrix. The reference extraction unit merges and combines the blue hue difference reference matrix and the red hue difference reference matrix in the matrix depth dimension to obtain the chromaticity reference data.

[0053] The feature extraction module is used to acquire motion image sequences of patients performing upper limb rehabilitation movements in real time, collected by the smart terminal. It constructs spatial skeleton vectors through the coordinates of key human nodes, extracts the current chromaticity data of the region of interest in the motion image sequence, and generates skin tone tension mutation data by comparing it with the chromaticity reference data.

[0054] Furthermore, the feature extraction module includes:

[0055] The skeleton tracking unit is used to extract the coordinates of key human body nodes in the current frame of the motion image sequence. It constructs the acromion vector by using the coordinates of the bilateral acromion nodes and constructs the midline vector of the spine by using the coordinates of the neck nodes and the midpoint of the pelvis. The acromion vector and the midline vector of the spine are used to form a spatial skeleton vector.

[0056] The dynamic chromaticity unit is used to dynamically lock the region of interest in the current frame based on the coordinates of key human body nodes, perform YCbCr color space mapping on the region of interest, strip the Y channel data, and obtain the current chromaticity data.

[0057] The tension calculation unit is used to calculate the distance deviation between the current chromaticity data and the chromaticity reference data to obtain skin tone tension mutation data;

[0058] The tension calculation unit is specifically used for:

[0059] Scaling the current chromaticity data and chromaticity reference data to a uniform preset resolution size;

[0060] For the current chromaticity data after resolution unification and the chromaticity reference data, calculate the squared deviation value in the Cb channel and the squared deviation value in the Cr channel respectively;

[0061] The square root of the squared deviation value under the Cb channel and the squared deviation value under the Cr channel is then used to obtain the skin tone tension mutation data.

[0062] Specifically, the skeleton tracking unit extracts the pixel matrix of the current frame from the motion image sequence and extracts the pixel coordinates of key human nodes in the current frame. The key node coordinates obtained specifically include the coordinates of the left acromion node, the right acromion node, the neck node, and the midpoint of the pelvis. The skeleton tracking unit subtracts the horizontal and vertical values ​​of the left acromion node coordinate from the horizontal and vertical values ​​of the right acromion node coordinate to calculate the horizontal and vertical displacement components, and combines them to form the acromion vector in a two-dimensional plane coordinate system. Simultaneously, the unit subtracts the horizontal and vertical values ​​of the neck node coordinate from the horizontal and vertical values ​​of the midpoint of the pelvis coordinate to calculate the horizontal and vertical components of the spinal displacement, and combines them to form the midline vector of the spine. The skeleton tracking unit packages the generated acromion vector and the midline vector of the spine into a spatial skeleton vector and writes the spatial skeleton vector into a shared memory register for downstream nodes to call.

[0063] Patient limb movement causes continuous physical displacement of the target observation area in the image coordinate system. The dynamic chromaticity unit obtains the coordinates of the acromion node and elbow joint node of the affected side in real time in the current frame. According to the spatial offset trajectory between the two coordinates, the pixel boundary from the axilla to the medial side of the biceps brachii on the affected side is recalculated and defined in the pixel array of the current frame. The region of interest image matrix of the current frame is dynamically extracted. The dynamic chromaticity unit performs color space mapping calculation on the extracted RGB format region of interest image matrix. According to the underlying color conversion mapping relationship, it is converted into a YCbCr color space multidimensional matrix pixel by pixel. The dynamic chromaticity unit directly removes the Y channel data matrix in the YCbCr color space multidimensional matrix, and retains the Cb channel pixel value matrix and Cr channel pixel value matrix after removing the Y channel to form the current chromaticity data. The current chromaticity data is then sent to the tension calculation unit.

[0064] The tension calculation unit receives the current chromaticity data and retrieves the chromaticity reference data pre-generated by the symmetry modeling module from the system database. Due to the changes in camera depth of field caused by limb movement, there is a mismatch in the number of rows and columns between the pixel matrix dimension corresponding to the current chromaticity data and the pixel matrix dimension corresponding to the chromaticity reference data. The tension calculation unit uses a bilinear interpolation image scaling algorithm to resample and align the Cb channel matrix and Cr channel matrix in the current chromaticity data, as well as the Cb channel matrix and Cr channel matrix in the chromaticity reference data, to a unified preset resolution size value in the spatial dimension.

[0065] For the multidimensional matrix after resolution and size unification, the tension calculation unit performs matrix deviation operations in the channel dimension. The unit subtracts the corresponding pixel values ​​from the Cb channel matrix of the current chroma data and calculates the sum of squares to obtain the squared deviation value for the blue chroma difference channel. Simultaneously, it subtracts the corresponding pixel values ​​from the Cr channel matrix of the current chroma data and calculates the sum of squares to obtain the squared deviation value for the red chroma difference channel. The tension calculation unit then sums the squared deviation values ​​for the blue and red chroma difference channels and performs a square root operation on the sum to finally calculate the skin tone tension mutation data, as shown in the following formula:

[0066] ;

[0067] in, Data representing skin tone tension variation. This represents the number of rows in the pixel matrix at a uniform preset resolution size. This represents the number of columns in the pixel matrix at a uniform preset resolution size. The blue difference channel matrix representing the current chroma data. Line number The pixel values ​​of the column. The blue difference channel matrix representing the chromaticity reference data. Line number The pixel values ​​of the column. The red hue difference channel matrix representing the current chroma data. Line number The pixel values ​​of the column. The red hue difference channel matrix representing the chromaticity reference data. Line number The pixel values ​​of the column.

[0068] The compensation identification module is used to calculate the scapular compensation angle based on the spatial skeletal vector, generate a dynamic tolerance threshold based on the edema index, and perform a dual-gated comparison with the skin tension mutation data and the scapular compensation angle to generate a compensation diagnosis signal.

[0069] Furthermore, the compensation identification module includes:

[0070] Angle calculation unit is used to calculate the degree of deviation of the angle between the acromion vector and the midline vector of the spine in the two-dimensional image plane in the spatial skeleton vector, as the scapular compensation angle.

[0071] The threshold generation unit is used to generate a dynamic tolerance threshold by adjusting the standard tolerance benchmark value according to the edema index through a preset logarithmic adjustment function.

[0072] The cross-validation unit is used to monitor whether the scapular compensation angle is greater than the dynamic tolerance threshold and whether the skin tension mutation data is greater than the pre-stored skin tension threshold within a preset time monitoring window.

[0073] The threshold generation unit and cross-validation unit are specifically used for:

[0074] When the edema index is determined to be less than or equal to zero, the edema index is forcibly truncated to zero, and the dynamic tolerance threshold is made equal to the standard tolerance benchmark value.

[0075] When the scapular compensation angle is determined to be greater than the dynamic tolerance threshold, and the skin tension mutation data is greater than the skin tension threshold, and it is confirmed that the two judgment conditions are triggered synchronously within the same monitoring window, a compensation diagnosis signal is generated.

[0076] Specifically, the angle calculation unit extracts the acromion vector and the spinal midline vector from the spatial skeletal vector. In the two-dimensional image plane coordinate system, the angle calculation unit extracts the horizontal and vertical coordinate components of the acromion vector and the spinal midline vector. The angle calculation unit calculates the dot product of these two vectors, divides it by the product of their magnitudes, and finally performs an inverse cosine operation on the ratio to calculate the angle deviation between these two spatial vectors in the two-dimensional plane. This angle deviation is defined as the scapular compensation angle, as shown in the following formula:

[0077] ;

[0078] in, Represents the compensatory angle of the scapula. Represents the shoulder peak vector in a two-dimensional planar coordinate system. This represents the vector of the spine midline in a two-dimensional plane coordinate system. The magnitude of the acromion vector. The magnitude of the vector representing the midline of the spine, and the calculated scapular compensation angle are sent to the cross-validation unit as the first-level conditional comparison parameter.

[0079] The threshold generation unit acquires edema index data stored in the system database. Before executing the logarithmic adjustment function calculation, the threshold generation unit first performs a numerical boundary check on the edema index. When the system logic determines that the read edema index value is less than or equal to zero, the threshold generation unit triggers a data truncation processing program, forcibly assigning the calculated value of the current edema index to zero, and controlling the calculation flow to skip the numerical amplification stage of the logarithmic function, directly making the output dynamic tolerance threshold value equal to the system's preset standard tolerance benchmark value. When the system logic determines that the edema index is strictly greater than zero, the threshold generation unit substitutes the edema index into the preset logarithmic adjustment function for calculation. The threshold generation unit multiplies the edema index by the preset first adjustment coefficient, adds one, extracts the natural logarithm, multiplies the logarithm by the preset second adjustment coefficient, and finally multiplies the superimposed calculation result by the standard tolerance benchmark value to generate the dynamic tolerance threshold, as shown in the following formula:

[0080] ;

[0081] In the formula, Represents the dynamic tolerance threshold. This represents the standard tolerance baseline value set at the system's underlying level. The input edema index, This represents the second preset adjustment coefficient. The first preset adjustment coefficient is represented by the dynamic tolerance threshold generated, which is sent to the memory buffer area of ​​the cross-validation unit along with the scapular compensation angle.

[0082] The cross-validation unit establishes a data cache queue containing timestamp attributes and sets time monitoring window parameters to constrain the judgment range. Within a single underlying operation cycle, the cross-validation unit synchronously receives the scapular compensation angle output by the angle calculation unit and the skin tension mutation data output by the feature extraction module. The cross-validation unit is configured with parallel first condition judgment program and second condition judgment program. The first condition judgment program extracts the value of the scapular compensation angle and compares it with the value of the dynamic tolerance threshold. When the value of the scapular compensation angle is strictly greater than the value of the dynamic tolerance threshold, the first trigger timestamp is recorded and the first condition is established status flag is output.

[0083] Simultaneously, the second condition determination program extracts the value of skin tension mutation data and compares it with the pre-stored skin tension threshold in the system configuration table. When the value of the skin tension mutation data is strictly greater than the skin tension threshold, a second trigger timestamp is recorded and a second condition fulfillment status flag is output. The cross-validation unit reads the output status flag information and associated timestamp values ​​of the first and second condition determination programs in real time. When the cross-validation unit confirms that the first and second condition determination programs are synchronously in a condition-fulfilled state, and the absolute time difference between the first and second trigger timestamps is less than or equal to the time monitoring window parameter, it confirms that a physical state of simultaneous bone displacement exceeding limits and subcutaneous tissue tension exceeding limits has occurred at the current moment, thereby generating a compensatory diagnostic signal containing information of the current time point. This compensatory diagnostic signal, as a trigger instruction for action correction, is pushed to the downstream prescription control module to execute subsequent blocking flow procedures.

[0084] The prescription control module is used to send an interrupt command to the smart terminal to stop the current rehabilitation training after receiving a compensatory diagnosis signal, issue a downgraded rehabilitation prescription, and issue a detection command after the downgraded intervention cycle ends to obtain the skin tension mutation data of the detection state, and perform adaptive flow control of the rehabilitation prescription based on the data comparison results;

[0085] Furthermore, the prescription control module is specifically used for:

[0086] After the duration of the downgraded rehabilitation prescription is reset to zero, a detection command is issued to require the patient to repeat the controlled test lifting motion.

[0087] The scheduling feature extraction module acquires the skin tone tension abrupt change data in the probe state under the probe state;

[0088] Compare the probe-state skin tension mutation data with the skin tension mutation data;

[0089] When the probed skin tension mutation data is less than the skin tension mutation data, the downgrade lock state is released and the prescription advanced instruction is regenerated.

[0090] When the detected skin tension mutation data is greater than or equal to the skin tension mutation data, the current rehabilitation training is forcibly terminated and an early warning message is generated;

[0091] Specifically, when a compensatory diagnosis signal containing a diagnosis identifier is parsed, the prescription control module generates an interrupt instruction data packet with a termination opcode in the instruction register and sends it to the network communication interface of the smart terminal. After receiving the instruction packet, the smart terminal clears the local video playback buffer, cuts off the output sequence of the current rehabilitation action video stream, and forcibly stops the current rehabilitation training. At the same time as triggering the interrupt instruction, the prescription control module accesses the local rehabilitation prescription relational database of the system, performs a search and comparison according to the preset low motion amplitude parameters, extracts the media source address of the soothing stretching type of action and the corresponding skeletal movement constraint parameter matrix, splices and packages the retrieved data into a downgraded rehabilitation prescription data packet, and sends it to the smart terminal.

[0092] When issuing the downgraded rehabilitation prescription data package, the prescription control module initializes a countdown variable in the system's timing register and writes the system's pre-configured downgrade intervention cycle duration value into this countdown variable. The prescription control module relies on the system's global clock signal to trigger a decrementing calculation program in each clock cycle, as shown in the following formula:

[0093] ;

[0094] in, Representing the The current value of the countdown variable for one clock cycle. Representing the Subtract the countdown variable value by one clock cycle. This represents the fixed step time value of the system's global clock signal. Representing the discrete time step sequence, the prescription control module continuously performs zero-value checks on the current value of the countdown variable. When the logic determines that the current value of the countdown variable decays to zero, i.e., the execution time of the downgraded rehabilitation prescription returns to zero, the prescription control module generates a probe instruction network request packet. This request packet encapsulates an array of coordinate guidance parameters for a single controlled lifting motion and an interface refresh control character. The prescription control module sends this probe instruction to the smart terminal, driving the smart terminal to render and generate a guided display interface containing the controlled test lifting motion.

[0095] After uploading the video stream data of the aforementioned test actions to the smart terminal, the prescription control module sends an operation scheduling control message with the current timestamp to the feature extraction module. Once scheduled, the feature extraction module captures the pixel matrix of the image in the detection state, performs color space conversion and matrix deviation calculation, generates skin tone tension mutation data in the detection state, and sends this data back to the comparison register of the prescription control module. The prescription control module reads the skin tone tension mutation data that previously triggered the system interrupt from the historical data stack. In the arithmetic logic unit, the prescription control module calculates the difference and extracts the sign bit between the detection state skin tone tension mutation data and the skin tone tension mutation data. Based on the polarity of the difference, it executes adaptive flow control of the rehabilitation prescription, as shown in the following formula:

[0096] ;

[0097] In the formula, This represents the next control state parameter output by the prescription control module. This represents the data on abrupt changes in skin tone tension during the detection state. Data representing skin tone tension variation. This parameter represents the parameter that unlocks the downgraded state. The global blocking state parameter representing the forced termination of training;

[0098] The prescription control module executes the corresponding branch control program based on the control state parameters output by the above logic. When the calculation output is the parameter to release the downgrade lock, it indicates that the value of the skin tension mutation data in the probe state is strictly less than the value of the skin tension mutation data. The prescription control module sets the lock flag in the system status register to a low level to release the downgrade lock state. Subsequently, the prescription control module re-extracts the advanced action parameter set based on the original training plan sequence, constructs a prescription advanced instruction containing advanced action features, and sends it to the smart terminal to restore the push of the normalized advanced rehabilitation training media stream. When the calculation output is the global blocking state parameter to force the end of training, it indicates that the value of the skin tension mutation data in the probe state is greater than or equal to the value of the skin tension mutation data. The prescription control module maintains the lock flag at a high level, sends a termination control code to the smart terminal to permanently cut off the media data stream, and forcibly ends the current rehabilitation training. Simultaneously, the prescription control module extracts the current device identification code, the skin tension mutation data in the probe state, and the skin tension mutation data, encapsulates them into a network warning information data packet according to the preset communication message format, and pushes the warning information to the receiving port of the doctor's control node through the communication protocol.

[0099] Example 2

[0100] This invention provides an internet-based rehabilitation and nursing method for breast cancer patients, comprising the following steps:

[0101] Images of the patient's upper limb at rest were acquired to calculate the edema index, and chromaticity reference data of the region of interest in the upper limb images were extracted.

[0102] The motion image sequence of the patient performing upper limb rehabilitation movements was acquired to construct a spatial skeletal vector, and skin tone tension variation data was generated by combining it with chromaticity reference data.

[0103] A dynamic tolerance threshold is generated based on the edema index. The spatial skeleton vector, skin tension mutation data and the dynamic tolerance threshold are then compared with the dynamic tolerance threshold under dual gating to generate a compensatory diagnostic signal.

[0104] Based on the compensated diagnosis signal, a downgraded rehabilitation prescription is issued to interrupt the current rehabilitation training. After the downgrade intervention, the probe state skin tension mutation data is obtained, and the adaptive flow control of the rehabilitation prescription is executed based on the data comparison results.

[0105] Specifically, the system first acquires images of the patient's upper limbs at rest, calculates the edema index, and simultaneously extracts chromaticity reference data of the region of interest in the upper limb images to establish the patient's underlying physiological reference state. When the patient begins rehabilitation training, the system acquires motion image sequences of the patient's upper limb rehabilitation movements in real time, and constructs spatial skeletal vectors reflecting postural displacement by parsing the sequences. At the same time, the system combines the previously extracted chromaticity reference data to calculate and generate skin tone tension variation data that characterizes the physical traction state of local tissues.

[0106] To accurately identify abnormal movements, the system adaptively generates a dynamic tolerance threshold based on a pre-calculated edema index. This threshold is then input into the computational logic along with real-time constructed spatial skeletal vectors and skin tension mutation data for dual-gating comparison. When the data features meet the comparison criteria, the system confirms compensatory behavior and generates a compensatory diagnosis signal. Based on this signal, the system immediately issues a downgraded rehabilitation prescription to safely interrupt the current rehabilitation training. After the downgrade intervention, the system re-initiates the test to obtain probe-state skin tension mutation data. Finally, the system strictly adheres to the comparison results of the probe-state skin tension mutation data to execute adaptive flow control of the rehabilitation prescription, thus completing a full closed loop from vital sign monitoring and abnormality blocking to prescription reset.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An internet-based rehabilitation and nursing system for breast cancer patients, characterized in that, Includes the following modules: The vital signs modeling module is used to acquire images of the patient's upper limbs in a resting state collected by a smart terminal, extract the length of the forearm bones and the pixel area of ​​the forearm, normalize and calculate the edema index, and perform color space conversion on the region of interest in the upper limb image to extract chromaticity reference data. The feature extraction module is used to acquire motion image sequences of patients performing upper limb rehabilitation movements in real time collected by the smart terminal, construct spatial skeleton vectors through the coordinates of key human nodes, extract the current chromaticity data of the region of interest in the motion image sequence, and generate skin tone tension mutation data by comparing with the chromaticity reference data. The compensation identification module is used to calculate the scapular compensation angle based on the spatial skeletal vector, generate a dynamic tolerance threshold based on the edema index, and perform a dual-gated comparison with the skin tension mutation data and the scapular compensation angle to generate a compensation diagnosis signal. The prescription control module is used to send an interrupt command to the smart terminal to stop the current rehabilitation training after receiving the compensatory diagnosis signal, issue a downgraded rehabilitation prescription, and issue a detection command to obtain the skin tension mutation data of the detection state after the downgrade intervention cycle ends, and perform adaptive flow control of the rehabilitation prescription based on the data comparison results.

2. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 1, characterized in that, The vital sign modeling module includes: The image acquisition unit is used to acquire bilateral upper limb images in a resting state, extract the coordinates of skeletal key points using a two-dimensional human pose estimation model, and locate the region of interest in the image of the affected upper limb based on the coordinates of the skeletal key points. The edema calculation unit is used to extract the pixel length of the bilateral forearm bones based on the coordinates of the skeletal key points, crop the bilateral forearm image regions based on the coordinates of the skeletal key points, and perform normalization calculation based on the pixel area of ​​the bilateral forearm image regions to obtain the edema index. The reference extraction unit is used to map the image data of the region of interest from the RGB color space to the YCbCr color space, strip the luminance channel data, and retain the chrominance channel data as the chrominance reference data.

3. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 2, characterized in that, The edema calculation unit is specifically used for: Extract the pixel length of the bones in the affected forearm and the pixel length of the bones in the healthy forearm; Extract the pixel area of ​​the affected forearm image region and the pixel area of ​​the healthy forearm image region from the bilateral forearm image regions; Divide the pixel area of ​​the affected forearm image region by the square of the pixel length of the affected forearm bone to obtain the normalized arm circumference feature value of the affected side. Divide the pixel area of ​​the healthy forearm image region by the square of the pixel length of the healthy forearm bone to obtain the normalized arm circumference feature value of the healthy side. The difference between the normalized arm circumference feature value of the affected side and the normalized arm circumference feature value of the healthy side is calculated, and the difference is divided by the normalized arm circumference feature value of the healthy side to obtain the edema index.

4. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 2, characterized in that, The reference extraction unit is specifically used for: The image data of the region of interest is mapped from the RGB color space to the YCbCr color space; Remove the Y channel data from the YCbCr color space; The Cb channel data is extracted to construct the blue hue difference benchmark matrix, and the Cr channel data is extracted to construct the red hue difference benchmark matrix. The chromaticity reference data is obtained by combining the blue hue difference reference matrix and the red hue difference reference matrix.

5. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 1, characterized in that, The feature extraction module includes: The skeleton tracking unit is used to extract the coordinates of key human body nodes in the current frame of the motion image sequence, construct an acromion vector through the coordinates of bilateral acromion nodes, construct a midline vector of the spine through the coordinates of the neck nodes and the midpoint of the pelvis, and use the acromion vector and the midline vector of the spine to form the spatial skeleton vector. The dynamic chromaticity unit is used to dynamically lock the region of interest in the current frame based on the coordinates of the key human body nodes, perform YCbCr color space mapping on the region of interest and strip the Y channel data to obtain the current chromaticity data; The tension calculation unit is used to calculate the distance deviation between the current chromaticity data and the chromaticity reference data to obtain the skin tone tension mutation data.

6. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 5, characterized in that, The tension calculation unit is specifically used for: The current chromaticity data and the chromaticity reference data are scaled to a uniform preset resolution size; For the current chromaticity data after resolution unification and the chromaticity reference data, calculate the squared deviation value in the Cb channel and the squared deviation value in the Cr channel respectively; The square root of the squared deviation value under the Cb channel and the squared deviation value under the Cr channel is then performed to obtain the skin tone tension mutation data.

7. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 1, characterized in that, The compensation identification module includes: An angle calculation unit is used to calculate the degree of deviation of the angle between the acromion vector and the midline vector of the spine in the two-dimensional image plane in the spatial skeletal vector, as the scapular compensation angle. The threshold generation unit is used to generate the dynamic tolerance threshold by adjusting the standard tolerance benchmark value according to the edema index through a preset logarithmic adjustment function. The cross-validation unit is used to monitor whether the scapular compensation angle is greater than the dynamic tolerance threshold and whether the skin tension mutation data is greater than the pre-stored skin tension threshold within a preset time monitoring window.

8. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 7, characterized in that, The threshold generation unit and the cross-validation unit are specifically used for: When the edema index is determined to be less than or equal to zero, the edema index is forcibly truncated to zero, and the dynamic tolerance threshold is made equal to the standard tolerance benchmark value. When it is determined that the scapular compensation angle is greater than the dynamic tolerance threshold, and the skin tone tension mutation data is greater than the skin tone tension threshold, and it is confirmed that the two judgment conditions are triggered synchronously within the same monitoring window, the compensation diagnosis signal is generated.

9. The Internet-based breast cancer patient rehabilitation and nursing system according to claim 1, characterized in that, The prescription control module is specifically used for: After the execution time of the downgraded rehabilitation prescription is reduced to zero, the detection command is issued to require the patient to re-perform the controlled test lifting action; The feature extraction module is scheduled to acquire the skin tone tension abrupt change data in the detection state under the detection state; Compare the skin tone tension mutation data of the probe state with the skin tone tension mutation data; When the detected skin tension mutation data is less than the skin tension mutation data, the downgrade lock state is released and the prescription advancement instruction is regenerated. When the detected skin tension mutation data is greater than or equal to the skin tension mutation data, the current rehabilitation training is forcibly terminated and an early warning message is generated.

10. An internet-based method for the rehabilitation and nursing care of breast cancer patients, characterized in that, The Internet-based breast cancer patient rehabilitation and care system according to any one of claims 1-9 includes the following steps: Images of the patient's upper limb at rest were acquired to calculate the edema index, and chromaticity reference data of the region of interest in the upper limb images were extracted. A sequence of motion images of the patient performing upper limb rehabilitation movements is acquired to construct a spatial skeletal vector, and skin tone tension mutation data is generated by combining the chromaticity reference data. A dynamic tolerance threshold is generated based on the edema index. The spatial skeleton vector, the skin tension mutation data, and the dynamic tolerance threshold are then compared under dual gating to generate a compensatory diagnostic signal. Based on the compensated diagnosis signal, a downgraded rehabilitation prescription is issued to interrupt the current rehabilitation training. After the downgrade intervention, the probe state skin tone tension mutation data is obtained, and adaptive flow control of the rehabilitation prescription is executed based on the data comparison results.

Images