Post-chemotherapy adverse reaction dynamic monitoring and follow-up management system

Abstract

The invention discloses an adverse reaction dynamic monitoring and follow-up management system after chemotherapy, and relates to the technical field of data processing, and the system comprises an acquisition module which determines a to-be-monitored symptom and a corresponding target part according to a chemotherapy regimen of a patient, and receives an image of the target part; the recognition module carries out registration on the part partition map and the target part image, a reaction partition with an abnormal color block area is determined, and color block positioning points are marked; the analysis module generates a positioning acquisition frame according to the part partition to which the color block positioning points belong, receives the review part image, cuts the target part image and the review part image based on the positioning acquisition frame, and determines the point location diffusion amount; and the decision-making module determines a follow-up visit mode according to the point location diffusion amount and the partition position of the reaction partition. The patient is guided to review and shoot the same position through the positioning collection frame, the follow-up visit mode is determined in combination with the partition positions, and the accuracy of follow-up visit decision is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy. Background Technology

[0002] Chemotherapy is a common treatment for cancer. Patients undergoing chemotherapy may experience various adverse reactions, among which skin-related adverse reactions are relatively common, such as hand-foot syndrome, rash, and injection site reactions. These skin adverse reactions usually appear gradually within a few days after the patient returns home from chemotherapy. Therefore, medical institutions need to follow up with patients to understand the occurrence and evolution of these adverse reactions.

[0003] Currently, follow-up monitoring of skin adverse reactions after chemotherapy mainly relies on telephone follow-ups or having patients upload photos themselves. Telephone follow-ups depend on patients' verbal descriptions of the lesions, but patients usually lack medical knowledge and find it difficult to accurately describe the location and severity of the lesions. While having patients upload photos provides visual information, the angle, distance, and framing of each photo may vary, making it difficult for doctors to determine whether multiple uploaded photos depict the same location. Even if the location is the same, there may be slight discrepancies, leading to misdiagnosis. Summary of the Invention

[0004] In view of the aforementioned problems, this application is hereby filed.

[0005] Therefore, this application provides a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, in order to solve the problem of how to achieve accurate tracking of the same lesion location across cycles in the scenario where chemotherapy patients take photos at home to monitor skin adverse reactions, and to make the tracking results have anatomical location significance to support clinical follow-up decisions.

[0006] To solve the above-mentioned technical problems, this application provides the following technical solution: In the first aspect, this application provides a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, including: a data acquisition module, used to determine the symptoms to be monitored and the corresponding target sites according to the patient's chemotherapy regimen, and to receive images of the target sites from the patient's terminal; The identification module is used to register the part partition map corresponding to the target part and the target part image, determine the part partition with abnormal color block area as the reaction partition, and mark the color block positioning point according to the position of the abnormal color block area; The analysis module is used to generate a positioning acquisition frame according to the part of the location to which the color block positioning point belongs, receive the follow-up image of the patient terminal according to the positioning acquisition frame, crop the target part image and the follow-up part image based on the positioning acquisition frame, and determine the point diffusion amount. The decision module is used to determine the follow-up method based on the diffusion amount of the point and the zoning position of the reaction zone in the site zoning map.

[0007] Preferably, the step of determining the symptoms to be monitored and the corresponding target sites based on the patient's chemotherapy regimen includes: Multiple candidate sites are identified based on the chemotherapy drug information in the chemotherapy regimen, and the site incidence rate of each candidate site is obtained; When the number of chemotherapy cycles in the chemotherapy regimen is determined to be the first cycle, the candidate site with the highest incidence rate is selected as the target site. When the number of chemotherapy cycles is determined to be a non-first cycle, candidate sites with reaction zones in previous chemotherapy cycles are retrieved as target sites.

[0008] Preferably, the registration of the region partition map corresponding to the target region and the target region image includes: Color block recognition is performed on the target area image, and abnormal color block areas in the target area image are marked as registration exclusion areas; After excluding the registration exclusion region in the target region image, the region contour of the target region is identified, and multiple contour feature points on the region contour are extracted; Obtain multiple boundary feature points on the outer boundary of the partition of the part partition map, calculate the contour feature spacing between adjacent contour feature points, and calculate the boundary feature spacing between adjacent boundary feature points. The region partition map is scaled based on the contour feature spacing and boundary feature spacing. The boundary feature points in the scaled region partition map are then aligned and superimposed with the contour feature points to complete the registration.

[0009] Preferably, the extraction of multiple contour feature points on the contour of the region includes: Calculate the average contour curvature of the part and the contour curvature of each contour point, and take the contour points whose contour curvature is less than the average contour curvature as candidate concavity positions. Obtain the boundary positions between adjacent part partitions in the part partitioning map, and count the number of the boundary positions as the number of anatomical depressions; Based on the number of candidate depression locations and the number of anatomical depressions, contour feature points are determined.

[0010] Preferably, determining the contour feature points based on the number of candidate depression locations and the number of anatomical depressions includes: When the number of candidate depression locations is equal to the number of anatomical depressions, the candidate depression locations are used as contour feature points. When the number of candidate depression locations is greater than the number of dissected depressions, calculate the positional distance between each candidate depression location and the boundary location of each partition, the candidate distance between adjacent candidate depression locations, and the average candidate distance. If the position spacing is determined to be the minimum value among the position spacings between each concave candidate position and the boundary position of the same partition, and the difference between the candidate spacing between the concave candidate position and the adjacent concave candidate position and the average value of the candidate spacing is less than the average value of the candidate spacing, then the concave candidate position is taken as a contour feature point. If the number of candidate depression locations is less than the number of anatomical depressions, a re-acquisition prompt is sent to the patient terminal, and the re-acquisition prompt includes a location expansion prompt.

[0011] Preferably, the area partitioning where abnormal color patches are identified is a reaction partition, including: The corresponding region of the registration exclusion region in the registered target region image is taken as the first region, and the region in each region partition other than the first region is taken as the second region. If the partition category of the part where the first region is located is determined to be a high-sensitivity partition, or if the partition category is a low-sensitivity partition and there is a marked abnormal color block area in the high-sensitivity partition, then the first region is marked as an abnormal color block area. When it is determined that the partition category is a low-sensitivity partition and there are no marked abnormal color block areas in the high-sensitivity partition, when it is determined that the first color difference between the first area and the second area is greater than the partition color difference between the second area of ​​the high-sensitivity partition and the low-sensitivity partition, the first area is marked as an abnormal color block area. The areas containing the abnormal color blocks are divided into reaction zones.

[0012] Preferably, marking color block positioning points based on the location of the abnormal color block area includes: When the number of abnormal color block regions is determined to be a single one, the center of the abnormal color block region is taken as the color block positioning point; When the number of abnormal color block regions is determined to be multiple, the center of the abnormal color block region with the largest area in the abnormal color block region located in the high-sensitivity partition is taken as the color block positioning point. If all abnormal color block regions are located in the low-sensitivity partition, the center of the abnormal color block region with the largest area is taken as the color block positioning point. When it is determined that the abnormal color block area spans adjacent part partitions, the part partition with the largest area of ​​the abnormal color block area within each part partition is taken as the reaction partition, and the center of the part of the abnormal color block area within the reaction partition is taken as the color block positioning point.

[0013] Preferably, the step of generating a positioning acquisition frame based on the region to which the color block positioning point belongs includes: The size of the acquisition frame is determined based on the area size of the abnormal color block region corresponding to the color block positioning point and the partition boundary of the part to which the color block positioning point belongs. A positioning acquisition frame is generated with the color block positioning point as the center and based on the acquisition frame size; The process of cropping the target area image and the review area image according to the positioning acquisition frame, and determining the point diffusion, includes: The initial positioning image is obtained by cropping the target area image according to the positioning acquisition frame, and the re-examination positioning image is obtained by cropping the re-examination area image according to the positioning acquisition frame. Calculate the first area of ​​the abnormal color block region in the initial positioning image and the second area of ​​the abnormal color block region in the re-examination positioning image, and determine the point diffusion amount based on the second area and the first area.

[0014] Preferably, determining the point diffusion amount based on the re-examination area and the initial area includes: obtaining the centroid offset direction and determining the partition category of adjacent parts partitions in the centroid offset direction based on the centroid of the first region of the abnormal color block area in the initial positioning image and the centroid of the second region of the abnormal color block area in the re-examination positioning image; When the partition category is determined to be a high-sensitivity partition, the point diffusion amount is determined based on the difference between the second area and the first area and the third area of ​​the positioning acquisition frame; When the partition category is determined to be a low-sensitivity partition, the point diffusion amount is determined based on the area difference between the second area and the first area, the third area, and the centroid offset distance between the centroid of the first region and the centroid of the second region.

[0015] Preferably, determining the follow-up method based on the point diffusion amount and the zoning position of the reaction zone in the site zoning map includes: When the diffusion amount at the point is determined to be positive and the reaction zone is classified as a highly sensitive zone, the follow-up method will be determined to be a return to the hospital for re-examination. When the point diffusion is determined to be positive and the partition category is low sensitivity partition, the follow-up method is to shorten the collection interval; When the diffusion amount at a given point is negative, and the absolute value of the diffusion amount is greater than the ratio of the first area to the third area, the follow-up method is determined to be extending the collection interval. When the absolute value of the diffusion amount at a given point is less than or equal to the ratio of the first area to the third area, the follow-up method is determined to be maintaining the collection interval.

[0016] Implementing this application will have the following beneficial effects: 1. This application registers a site zoning map with images of the target site taken from the patient. In the registered image, site zoning areas containing abnormal color patches are identified as reaction zoning areas, allowing the identified abnormal areas to correspond to specific anatomical zoning areas. Since different anatomical zoning areas have varying degrees of impact on patients' daily functions—for example, the finger zoning area of ​​the palm has a greater impact on grasping function than the palm zoning area—this application considers both the spread of the lesions and the location of the reaction zoning area on the site zoning map when determining the follow-up method. This allows for follow-up arrangements that match the clinical significance of lesions in different zoning areas, thereby improving the accuracy of follow-up decisions.

[0017] 2. This application generates a location acquisition frame based on the location of the color block positioning point obtained in the initial identification. When the patient undergoes a follow-up examination, the location acquisition frame is sent to the patient's terminal, guiding the patient to take images of the follow-up area according to the frame. The purpose is to ensure that the patient can capture the same area during the follow-up examination as in the initial acquisition, solving the problem of inconsistent image positions in multiple photos. Based on the location acquisition frame, the target area image and the follow-up area image are cropped to obtain the initial location image and the follow-up location image. By comparing the area changes of abnormal color block regions in the two location images, the diffusion amount of the point can be determined, allowing for accurate quantification of the spread or regression of the lesion and providing a reliable basis for follow-up decisions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall structure of a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, which is involved in this application. Figure 2 This is a zoning diagram of the palm area in a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, which is involved in this application. Figure 3This is a schematic diagram of the contour feature point extraction of a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy involved in this application; Figure 4 This is a diagram of a computer device for a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, which is the subject of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In this field, patients typically return home to recuperate after completing chemotherapy in the hospital. However, skin adverse reactions caused by chemotherapy drugs often only gradually appear within days of discharge, such as hand-foot syndrome and rashes. Doctors cannot directly observe these changes on the patient's skin and usually ask the patient to send photos. However, patients are not professionals, and the angles and distances of the photos vary each time. Therefore, it is difficult for doctors to determine whether the photos taken this time are of the same location as those taken last time, making it hard to say whether the condition is improving or worsening.

[0022] To address the aforementioned issues, this application proposes a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, such as... Figure 1 As shown, it includes a data acquisition module, an identification module, an analysis module, and a decision-making module.

[0023] The core problems addressed in this application are: first, to ensure that identified abnormal color patches can be mapped to specific anatomical regions; second, to ensure that two consecutive photographs of the patient capture the same location; and third, to quantify changes in the patient's condition using a numerical value. The corresponding solution is as follows: First, register the patient's photograph with a standard anatomical region map to determine whether the abnormal color patch is on the finger or palm. Then, mark a location point at the location of the abnormal color patch, and generate a capture frame based on this point. Subsequent photographs of the patient will follow this frame. Finally, crop and compare the two images to calculate the change in the area of ​​the abnormal color patch as the point diffusion rate. This, combined with the patient's location within the anatomical region, will determine the next follow-up strategy.

[0024] The acquisition module determines the symptoms to be monitored and the corresponding target sites based on the patient's chemotherapy regimen, and receives images of the target sites from the patient's terminal.

[0025] It's important to note that in home monitoring scenarios for adverse skin reactions after chemotherapy, patients often have limited medical knowledge and are unsure which parts of their body to focus on. Furthermore, the same chemotherapy drug can affect multiple sites; for example, patients using capecitabine may experience reactions in their palms, soles, and oral mucosa. Given limited patient compliance, it's impractical to have patients photograph all potentially affected areas at each follow-up visit. Therefore, a priority site for data collection needs to be identified. The data collection module in this application automatically determines the most critical target site for monitoring based on the patient's ongoing chemotherapy regimen, guiding the patient to take photographs.

[0026] Specifically, the "determining the symptoms to be monitored and the corresponding target sites based on the patient's chemotherapy regimen" in the data acquisition module includes steps A1 to A3: Step A1: Identify multiple candidate sites based on the chemotherapy drug information in the chemotherapy regimen, and obtain the site incidence rate for each candidate site.

[0027] Candidate sites refer to the body parts where chemotherapy drugs in a chemotherapy regimen may cause skin reactions, and site incidence rate refers to the proportion of patients using the drug who experience a reaction at the corresponding candidate site. These data are derived from drug instructions or clinical literature.

[0028] Understandably, the data acquisition module extracts chemotherapy drug information from the chemotherapy regimen, queries the pre-stored correspondence between drugs and reaction sites, and obtains multiple candidate sites and their corresponding incidence rates.

[0029] For example, if a patient's chemotherapy regimen includes capecitabine, common adverse reactions are hand-foot syndrome and oral mucositis. Therefore, the data collection module will obtain multiple candidate sites such as the palms, soles, and mouth after querying.

[0030] Step A2: When the number of chemotherapy cycles in the chemotherapy regimen is determined to be the first cycle, select the candidate site with the highest incidence rate as the target site.

[0031] It should be noted that since chemotherapy is usually performed in multiple cycles, if there is no historical monitoring data for the patient during the first cycle, the data collection module will select the site of occurrence based on the population statistics. The site that is most likely to cause problems will be monitored first, and the candidate site with the highest site of occurrence will be selected as the target site.

[0032] The logic behind selecting the candidate site with the highest incidence rate as the target site is that since this site is most likely to cause a reaction in patients using similar drugs, prioritizing monitoring this site for a new patient is also the most likely way to detect problems. Continuing the previous example, if the patient is receiving capecitabine chemotherapy for the first time, the acquisition module will select the palm as the target site because the palm is a typical site for hand-foot syndrome, and the incidence rate is higher than that of the oral cavity.

[0033] Step A3: When the number of chemotherapy cycles is determined to be a non-first cycle, retrieve the candidate sites with reaction zones in the previous chemotherapy cycles as the target sites.

[0034] It's important to note that chemotherapy-induced skin reactions have a unique characteristic: if a certain area has shown a reaction in a previous cycle, it is more likely to relapse or worsen in subsequent cycles. Therefore, during non-first cycles, the data acquisition module will prioritize candidate sites where a reaction zone has already appeared in previous cycles, retrieving candidate sites with reaction zones from previous chemotherapy cycles as target sites, and continuously tracking the evolution of the same site.

[0035] Furthermore, the "presence of a reaction zone" mentioned here refers to the identification module recognizing abnormal color patches in the image of a candidate site during the monitoring of previous chemotherapy cycles, and marking the site with the abnormal color patches as a reaction zone. If a candidate site has previously shown a reaction zone in a previous cycle, then that candidate site will be prioritized as the target site for the current cycle.

[0036] For example, suppose that during the first cycle, the patient selected the palm as the target site based on the incidence rate and monitored it, identifying abnormal color patches in the finger areas of the palm. Then, when the patient enters the second cycle, the acquisition module will retrieve the palm—a candidate site that showed a reaction in the previous cycle—as the target site, instead of reselecting based on the incidence rate. The advantage of this is that it allows for continuous tracking of the same site that has already shown a reaction, observing its evolution across multiple chemotherapy cycles.

[0037] Preferably, steps A1 to A3 address individual variability in monitoring chemotherapy-induced skin adverse reactions by employing different target site determination methods for the first and non-first cycles. In the first cycle, due to a lack of historical patient data, selecting the most likely site for reaction based on population statistics of site incidence is a reasonable cold-start approach. In non-first cycles, however, actual reaction data from previous cycles is fully utilized, focusing on sites where reactions have already occurred. This ensures that target site selection is both data-driven when individual data is lacking and allows for more targeted tracking once individual data is accumulated, aligning with the clinical reality of multiple chemotherapy cycles and the tendency for adverse reactions to recur.

[0038] Ideally, the acquisition module automatically identifies the target area and guides the patient to acquire images of that area, reducing the requirement for the patient's medical knowledge. Patients simply need to take photos of the designated area and upload them as prompted, without needing to determine which area should be monitored. This reduces the patient's workload, improves follow-up compliance, and avoids the possibility of missing important monitoring areas due to patient misjudgment.

[0039] The recognition module registers the part partition map corresponding to the target part with the target part image, determines the part partition with abnormal color block area as the reaction partition, and marks the color block positioning point according to the position of the abnormal color block area.

[0040] It's important to note that the image of the target area taken by the patient is just a photograph. The doctor sees a red spot in the photo, but doesn't know its exact location on the palm—whether it's on a finger or in the center. The impact of reactions in different locations on a patient's daily function varies; problems in the fingers affect grasping, while problems in the palm have a relatively smaller impact. Therefore, the recognition module's task is to align and overlay the patient's photograph with a standard area partitioning map, thus determining which partition the red spot falls into.

[0041] like Figure 2 As shown, the site zoning diagram is a pre-established standard anatomical zoning template. Taking the palm as an example, the site zoning diagram is divided into finger zoning, palm zoning, thenar zoning, and hypothenar zoning according to anatomical functional areas. Figure 2The solid black dots represent the boundaries between different areas, that is, the points where the boundaries of adjacent areas intersect on the outer boundary of the area. For the palm, this means the spaces between the fingers. The thin solid lines in the image represent the boundaries of different areas, used to distinguish them. Different areas have varying degrees of impact on the patient's daily function. For example, the finger areas directly affect grasping function and are considered highly sensitive areas; the palm areas have a relatively smaller impact and are considered low-sensitivity areas. During registration, the recognition module aligns and overlays the image of the target area captured by the patient with the area partition map, ensuring that any abnormal color patches are mapped to specific areas.

[0042] Specifically, the execution steps of the identification module include steps B1 to B3: Step B1: Register the part partition map corresponding to the target part and the target part image.

[0043] It should be noted that the image of the target area taken by the patient with their mobile phone and the pre-stored area partition image are different in size and angle, and need to be aligned and superimposed through registration. However, there is a difficulty in registration: if there are abnormal color patches such as erythema in the target area image, the color and shape of these areas have changed, which will interfere with contour recognition. Therefore, the abnormal color patch areas must be excluded before registration.

[0044] Specifically, step B1 includes steps B11 to B14: Step B11: Perform color block recognition on the target area image and mark abnormal color block areas in the target area image as registration exclusion areas.

[0045] Understandably, the recognition module first performs color block recognition on the target area image, identifying regions where the color significantly deviates from normal skin tone and marking them as registration exclusion areas. When extracting contour feature points, these registration exclusion areas are skipped to avoid abnormal color blocks interfering with registration.

[0046] Step B12: After excluding the registration exclusion area in the target area image, identify the contour of the target area and extract multiple contour feature points on the contour.

[0047] It should be noted that registration requires finding the correspondence between the target area image and the area partitioning image, and contour feature points are the anchor points for establishing this correspondence. Taking the palm as an example, the spaces between the fingers are excellent contour feature points because the spaces between the fingers appear as indentations on the contour, their positions are stable and easy to identify, and the spaces between the fingers are also the boundaries between adjacent area partitions in the area partitioning image.

[0048] Furthermore, step B12, "extracting multiple contour feature points on the part contour," includes steps B121 to B123: Step B121: Calculate the average contour curvature of the part and the contour curvature of each contour point, and select contour points with contour curvature less than the average contour curvature as candidate concave locations.

[0049] like Figure 3 As shown, when extracting contour feature points, the recognition module first performs curvature analysis on the contour of the target area in the image, calculating the contour curvature of each contour point. Contour curvature describes the degree of curvature of a contour line at a point; the contour curvature is larger at convex locations and smaller at concave locations. The hollow circles in the figure represent candidate concave locations selected through curvature analysis, i.e., contour points with a contour curvature less than the average contour curvature. For the palm, these are distributed in contour depressions such as between the fingers, knuckle creases, and joint wrinkles. Figure 3 The solid circle represents the final determined contour feature point, which is obtained by filtering from the candidate depression positions after matching with the boundary positions of the partition map. When the number of candidate depression positions is the same as the number of anatomical depressions, the candidate depression positions are the contour feature points; when the number of candidate depression positions is greater than the number of anatomical depressions, it is necessary to filter by position spacing and candidate spacing to eliminate misidentification caused by factors such as skin folds.

[0050] Understandably, the recognition module calculates the average contour curvature of the area's outline, then compares it point by point, filtering out contour points with curvature less than the average contour curvature as candidate depression locations. These candidate depression locations could be genuine anatomical depressions, such as between fingers, or they could be misidentified due to skin folds or shadows.

[0051] Step B122: Obtain the boundary positions between adjacent part partitions in the part partition map, and count the number of partition boundary positions as the number of anatomical depressions.

[0052] In this context, the boundary position of a region is the intersection point of the boundaries of adjacent regions on the outer boundary of the region in the region-specific map. For the palm, this means the location of each finger gap. The number of anatomical depressions represents the number of boundary positions of regions, indicating the theoretically identifiable number of depression feature points.

[0053] Step B123: Determine the contour feature points based on the number of candidate depression locations and the number of anatomical depressions.

[0054] It should be noted that when the patient's fingers are taken, they may not be fully open, or there may be skin folds. The number of candidate indentation locations identified may not match the number of anatomical indentations. If the number of candidate indentation locations is greater than the number of anatomical indentations, it indicates a false identification that needs to be filtered. If the number of candidate indentation locations is less than the number of anatomical indentations, it indicates that some finger gaps are obscured, and the patient should be prompted to take a new photo.

[0055] Preferably, steps B121 to B123 identify candidate locations for depressions through curvature analysis, and then verify the number of anatomical depressions in the location partition map, so that the extraction of contour feature points can not only adapt to the differences in hand shape among different patients, but also filter out misidentifications caused by skin folds.

[0056] Furthermore, step B123 includes steps B1231 to B1234: Step B1231: When the number of candidate depression locations is equal to the number of dissected depressions, the candidate depression locations are used as contour feature points.

[0057] It is easy to understand that the number of candidate depression locations is exactly equal to the number of anatomical depressions, indicating that the recognition results match the anatomical structure, and the candidate depression locations are directly used as contour feature points.

[0058] Step B1232: When the number of candidate depression locations is greater than the number of dissected depressions, calculate the positional distance between each candidate depression location and the boundary location of each partition, the candidate distance between adjacent candidate depression locations, and the average candidate distance.

[0059] It should be noted that the number of candidate depression locations exceeds the number of anatomical depressions, indicating false positives. Therefore, it is necessary to filter out the true finger gap locations. True anatomical depressions have two characteristics: they are located near the boundaries of zones on the anatomical map; and the spacing between adjacent depressions is relatively uniform, as the width of a person's finger does not vary significantly.

[0060] Step B1233: Determine the position spacing as the minimum value among the position spacings between each concave candidate position and the boundary position of the same partition, and when the difference between the candidate spacing between the concave candidate position and the adjacent concave candidate position and the average candidate spacing is less than the average candidate spacing, then take the concave candidate position as the contour feature point.

[0061] Understandably, the recognition module performs two checks on each candidate depression location: the first is a position check, to see if the candidate depression location is the closest to a certain partition boundary; the second is a spacing check, to see if the candidate spacing between the candidate depression location and its adjacent candidate depression locations is close to the average candidate spacing. Only depression candidate locations that pass both checks are determined as contour feature points.

[0062] In one specific implementation, candidate depression locations closest to the boundary of the partition are selected based on positional spacing. However, if there happens to be a skin fold near the actual finger gap, the fold may be closer to the boundary than the actual finger gap. Relying solely on positional spacing and spacing uniformity may not be sufficient; verification from the curvature dimension is also necessary. Another characteristic of genuine anatomical depressions is that the degree of depression in each finger gap is similar, and the contour curvature values ​​should be relatively close. In contrast, misidentified depressions, such as skin folds or shadows, may have significantly larger or smaller curvatures, differing considerably from the curvature of the actual finger gaps.

[0063] Based on the above considerations, when the number of candidate depression locations is greater than the number of anatomical depressions, the process of selecting contour feature points from the candidate depression locations includes: Calculate the positional distance between each candidate depression position and the boundary position of each partition. For each partition boundary position, select the candidate depression position with the smallest positional distance from the partition boundary position as the initial screening depression position. Calculate the initial screening distance between adjacent initial screening depressions, and calculate the average initial screening distance for each initial screening distance. Calculate the mean curvature of the profile curvature corresponding to each initial screening depression position; If the difference between the initial screening distance and the average initial screening distance between the initial screening depression position and the adjacent initial screening depression position is less than the average initial screening distance, and the difference between the contour curvature corresponding to the initial screening depression position and the average curvature is less than the average curvature, then the initial screening depression position is taken as the contour feature point. If the difference between the initial screening distance and the average initial screening distance between the initial screening depression position and the adjacent initial screening depression position is greater than or equal to the average initial screening distance, or if the difference between the contour curvature corresponding to the initial screening depression position and the average curvature is greater than or equal to the average curvature, then the initial screening depression position is marked as the position to be verified. From the remaining depression candidate positions, the second smallest depression candidate position with the position distance at the boundary of the partition corresponding to the position to be verified is selected to replace the position to be verified, and the initial screening distance and contour curvature are verified again.

[0064] Understandably, the recognition module first filters out initial screening depression locations based on positional spacing. Then, it performs two checks on these locations: a spacing check to see if the initial screening distance between the depression location and its adjacent locations is close to the average initial screening distance; and a curvature check to see if the contour curvature of the depression location is close to the average curvature. Only depression locations that pass both checks are identified as contour feature points. If a depression location fails the checks, the recognition module selects the next smallest positional spacing from the remaining candidate depression locations and replaces it, then re-checks until a suitable contour feature point is found.

[0065] Step B1234: When the number of candidate depression locations is less than the number of anatomical depressions, a re-acquisition prompt is sent to the patient terminal. The re-acquisition prompt includes a location expansion prompt.

[0066] Understandably, the number of candidate indentation locations is less than the number of anatomical indentations, indicating that the patient's fingers were not fully open when the image was taken, and part of the finger gaps were obscured. The recognition module sends a re-acquisition prompt to the patient's terminal, including a prompt to open the area, guiding the patient to open their fingers and retake the image.

[0067] Preferably, step B123 designs processing methods for three different relationships between the number of candidate depression locations and the number of anatomical depressions: when the number is equal, it is directly confirmed; when the number is too large, it is screened through multiple dimensions such as location spacing, candidate spacing and contour curvature; when the number is too small, it prompts for re-acquisition, so that the registration process can adapt to the situation where the patient's shooting posture is not standard.

[0068] Step B13: Obtain multiple boundary feature points on the outer boundary of the partition map, calculate the contour feature spacing between adjacent contour feature points, and calculate the boundary feature spacing between adjacent boundary feature points.

[0069] In this context, boundary feature points are the points on the outer boundary of a region's partition image that correspond to the intersection of the partitions. For a hand region partition image, these are the positions of the finger seams on the outer boundary. Contour feature spacing is the distance between adjacent contour feature points in the target region image, while boundary feature spacing is the distance between adjacent boundary feature points in the region partition image.

[0070] Step B14: Scale the part partition map based on the contour feature spacing and boundary feature spacing, and align and superimpose the boundary feature points and contour feature points in the scaled part partition map to complete the registration.

[0071] Understandably, due to varying patient shooting distances, the size of the hand in the target area image often differs from the size of the hand in the area partitioning image. The recognition module scales the area partitioning image based on the ratio of contour feature spacing to boundary feature spacing, ensuring the scaled area partitioning image matches the hand size in the target area image. Then, it aligns and superimposes the boundary feature points with the contour feature points to complete the registration. After registration, each pixel in the target area image corresponds to a specific area partition in the area partitioning image.

[0072] Preferably, step B1 first excludes abnormal color block areas and then extracts contour feature points, avoiding interference from chemotherapy skin reactions on registration; by matching and scaling feature points based on anatomical depressions, the registration of the target area image and the area partition map is completed, so that the identified abnormal color blocks can correspond to specific anatomical partitions.

[0073] Step B2: Determine the areas containing abnormal color patches as reaction zones.

[0074] It should be noted that patients' skin may have inherent characteristics such as skin color differences, birthmarks, and age spots. How can these inherent characteristics be distinguished from the abnormal color patches caused by chemotherapy? If historical data is available for comparison, the patient's original skin characteristics were present at the time of the initial data collection, while the abnormal color patches caused by chemotherapy appeared or worsened after chemotherapy. This comparison would allow for differentiation. However, the logic of this application is to identify the abnormal color patches first and then generate the localization acquisition frame. Therefore, there is no historical data available for comparison at the time of the initial data collection.

[0075] So, from another perspective, since there's no historical data during the initial image acquisition, how do we distinguish abnormal color patches within a single image? Chemotherapy-induced skin reactions typically appear in specific predisposing areas, such as the palmar surface of the fingers or pressure points on the palm. Therefore, we can use the predisposition of each area in the site zoning map to aid in the judgment. If a color patch is located in a highly sensitive zone, it is more likely to be an abnormal color patch caused by chemotherapy.

[0076] Specifically, step B2 includes steps B21 to B24: Step B21: Take the corresponding region of the registration exclusion region in the registered target region image as the first region, and take the region other than the first region in each region partition as the second region.

[0077] Understandably, after registration is complete, the positions of the previously marked registration exclusion areas in the target area image can correspond to specific area partitions. The recognition module uses the registration exclusion areas as the first region to be judged, and the areas within each area partition other than the first region as the second region. By comparing the color differences between the first and second regions, it determines whether the first region is an abnormal color patch area caused by chemotherapy.

[0078] Step B22: If the partition category of the part where the first region is located is a high-sensitivity partition, or the partition category is a low-sensitivity partition and there is a marked abnormal color block area in the high-sensitivity partition, then mark the first region as an abnormal color block area.

[0079] It's important to note that highly sensitive regions are common sites for chemotherapy-induced skin reactions, such as the finger areas of the palm. The first region is located in a highly sensitive region; this patch is very likely caused by chemotherapy and should be directly marked as an abnormal patch area. Conversely, if the first region is located in a low-sensitivity region, but an abnormal patch area already exists within a highly sensitive region of the same target area, it indicates that the patient has indeed experienced a chemotherapy-induced skin reaction. The patch in the low-sensitivity region is also highly likely to be caused by chemotherapy and should also be marked as an abnormal patch area.

[0080] Step B23: When the partition category is determined to be a low-sensitivity partition and there are no marked abnormal color block areas in the high-sensitivity partition, if the first color difference between the first area and the second area is greater than the partition color difference between the second area of ​​the high-sensitivity partition and the low-sensitivity partition, the first area is marked as an abnormal color block area.

[0081] It should be noted that since the first region is located in a low-sensitivity zone and no abnormal color patches were found in the high-sensitivity zone, it is necessary to more carefully determine whether this color patch is caused by chemotherapy or is an inherent characteristic of the patient's skin. The recognition module calculates the first color difference between the first region and the second region within the same zone, and simultaneously calculates the zone color difference between the second region within the high-sensitivity zone and the second region within the low-sensitivity zone. The zone color difference reflects the natural color difference between different parts of the patient's skin. If the first color difference is greater than the zone color difference, it indicates that the color abnormality of the first region exceeds the range of natural skin color differences, and the first region is marked as an abnormal color patch area.

[0082] Step B24: Divide the areas containing abnormal color patches into reaction zones.

[0083] Understandably, after completing the above judgment, the recognition module will mark the areas with abnormal color blocks as reaction zones.

[0084] Preferably, step B2 combines the sensitivity of the site zoning to judge abnormal color patches. High-sensitivity zoning uses a lenient standard, while low-sensitivity zoning uses a lenient standard when there are abnormalities in high-sensitivity zoning and a strict color difference comparison standard when there are no abnormalities in high-sensitivity zoning. This can capture skin reactions caused by chemotherapy and filter out the inherent pigmentation characteristics of the patient's skin.

[0085] Step B3: Mark the location points of the abnormal color blocks according to their positions.

[0086] It should be noted that the location point of the color patch needs to be able to represent the location of the abnormal color patch area, but there are several special cases to consider: chemotherapy-induced erythema may be irregular in shape, and the geometric center may not be inside the color patch; abnormal color patches may cross the partition boundary, and a single erythema may cover two parts of the partition at the same time. The location point should be marked on which part of the partition; there may also be multiple abnormal color patches at the same time, and it is necessary to determine which color patch's location point to mark.

[0087] Specifically, step B3 includes steps B31 to B33: Step B31: When the number of abnormal color block areas is determined to be a single one, the center of the abnormal color block area is taken as the color block positioning point.

[0088] It's easy to understand that if there's only one abnormal color patch area in the target area image, the center of that abnormal color patch area can be directly used as the color patch location point.

[0089] Step B32: When there are multiple abnormal color block regions, obtain the partition category of the reaction partition where each abnormal color block region is located. Take the center of the largest abnormal color block region in the high-sensitivity partition as the color block positioning point. If all abnormal color block regions are located in the low-sensitivity partition, take the center of the largest abnormal color block region as the color block positioning point.

[0090] It should be noted that when multiple abnormal color patches exist simultaneously, it is necessary to determine which one should be the primary monitoring target. The response in highly sensitive zones has a greater impact on patient function, so abnormal color patch areas located within highly sensitive zones should be prioritized. Among these, the largest abnormal color patch area should be selected, and its center should be used as the color patch location point. If all abnormal color patch areas are located within low-sensitivity zones, selection should be based on area, with the center of the largest abnormal color patch area used as the color patch location point.

[0091] Step B33: When it is determined that the abnormal color block area crosses an adjacent part zone, the part zone with the largest area of ​​the abnormal color block area within each part zone shall be taken as the reaction zone, and the center of the part of the abnormal color block area within the reaction zone shall be taken as the color block positioning point.

[0092] It should be noted that chemotherapy-induced skin reactions sometimes present as a single erythema spanning two adjacent areas, such as extending from the finger area to the palm area. The recognition module calculates the area within each area of ​​the abnormal erythema, designating the area with the largest area as the reaction area. Simultaneously, the center of the portion of the abnormal erythema within the reaction area is used as the erythema location point, ensuring that the location point falls within the reaction area.

[0093] Preferably, step B3 designs marking rules for color block positioning points for three cases: single abnormal color block, multiple abnormal color blocks, and abnormal color blocks that span partitions. When there are multiple abnormal color blocks, the abnormal color blocks in the high-sensitivity partitions are selected first. When they span partitions, the main partition is determined by the area within the partition, which is consistent with the logic of clinically judging the location of the main lesion.

[0094] Preferably, the recognition module aligns the target area image with the area partition map through registration, so that the identified abnormal color block areas can correspond to specific anatomical partitions; by combining the partition sensitivity to judge abnormalities, it distinguishes between chemotherapy reactions and inherent skin features; by marking color block positioning points in abnormal color block areas, it provides a positional reference for the generation of positioning acquisition frames.

[0095] The analysis module generates a positioning acquisition frame based on the location of the color block positioning point, receives the follow-up images of the area captured by the patient terminal according to the positioning acquisition frame, crops the target area image and the follow-up area image based on the positioning acquisition frame, and determines the point diffusion.

[0096] It should be noted that while the recognition module has already marked color-coded location points in the target area image, for monitoring chemotherapy-induced skin adverse reactions, it is more important to track changes in the lesion between multiple image acquisitions. The analysis module in this application generates a location acquisition frame based on the initial recognition result, guiding the patient to capture the same location during subsequent follow-up examinations. Then, by comparing the two images, the spread or regression of the lesion can be quantified.

[0097] Specifically, the execution steps of the analysis module include steps C1 to C3: Step C1: Generate a positioning acquisition frame by dividing the area according to the location of the color block positioning point.

[0098] It's important to note that a positioning acquisition frame that's too small may not be able to completely cover the abnormal color patch area, as some of the lesion will fall outside the frame after it spreads. Conversely, a positioning acquisition frame that's too large will include too much irrelevant normal skin area, affecting the accuracy of subsequent comparisons. Therefore, the size of the positioning acquisition frame should be determined based on the actual size of the abnormal color patch area, while allowing sufficient margin at the edges to accommodate lesion spread.

[0099] Furthermore, step C1 includes steps C11 and C12: Step C11: Determine the size of the acquisition frame based on the area size of the abnormal color block region corresponding to the color block positioning point and the partition boundary of the part to which the color block positioning point belongs.

[0100] Understandably, the analysis module obtains the area size of the abnormal color patch and calculates the initial acquisition frame size using a certain margin coefficient. For example, if the area size of the abnormal color patch is 20 mm long and 15 mm wide, the analysis module calculates the initial acquisition frame size as 30 mm long and 22.5 mm wide using a margin coefficient of 1.5. Furthermore, if the abnormal color patch area is close to the partition boundary, the acquisition frame size is also limited by the distance from the color patch positioning point to the partition boundary. For example, if the distance from the color patch positioning point to the nearest partition boundary is 25 mm, then the acquisition frame size in the corresponding direction cannot exceed 50 mm to avoid the positioning acquisition frame spanning multiple partitions.

[0101] Step C12: Generate a positioning acquisition frame centered on the color block positioning point and based on the acquisition frame size.

[0102] Understandably, the analysis module uses the color block positioning point as the geometric center of the positioning acquisition frame, generates a rectangular positioning acquisition frame according to the acquisition frame size determined in step C11, and sends it to the patient terminal for display on the shooting interface.

[0103] Preferably, step C1 determines the size of the acquisition frame based on the area size of the abnormal color block region and combines it with the partition boundary for constraint, so that the positioning acquisition frame can completely cover the abnormal color block region and its possible diffusion range, without including too many irrelevant areas.

[0104] Step C2: Receive images of the re-examination area taken by the patient's terminal according to the positioning acquisition frame.

[0105] Understandably, if patients follow the guidance of the positioning and acquisition frame on the interface to take pictures, it can be ensured that the images of the re-examination sites include the same area as the first acquisition.

[0106] Step C3: Based on the positioning acquisition frame, crop the target area image and the review area image respectively, and determine the point diffusion amount.

[0107] It should be noted that the distance of the phone from the patient during the two imaging sessions may differ, resulting in variations in the size of the palm in the images. Simply calculating the area difference as the extent of lesion diffusion is insufficient. Firstly, different imaging distances can lead to errors in area measurement. Secondly, even with the same area diffusion, the clinical significance differs depending on whether the lesion spreads to a small or large area. Furthermore, the impact on patient function differs depending on whether the lesion spreads to a highly sensitive or low-sensitivity area. Therefore, the calculation of lesion diffusion requires normalization and consideration of the clinical significance of the diffusion direction.

[0108] Specifically, step C3 includes steps C31 and C32: Step C31: Obtain the initial positioning image by cropping the target area image from the positioning acquisition frame, and obtain the re-examination positioning image by cropping the re-examination area image from the positioning acquisition frame.

[0109] Understandably, since the same positioning acquisition frame was used for both cropping operations, the initial positioning image and the re-examination positioning image correspond in anatomical position and can be used for area comparison.

[0110] Step C32: Calculate the first area of ​​the abnormal color block region in the initial positioning image and the second area of ​​the abnormal color block region in the re-examination positioning image, and determine the point diffusion amount based on the second area and the first area.

[0111] Understandably, the analysis module performs color patch recognition on both the initial localization image and the follow-up localization image, calculating the area of ​​the abnormal color patch region for each. However, the area difference alone cannot accurately reflect the actual changes in the lesion; a comprehensive judgment is also needed, taking into account the direction of diffusion.

[0112] Furthermore, step C32, "determining the point diffusion amount based on the second area and the first area," includes steps C321 to C323: Step C321: Based on the centroid of the first region of the abnormal color block area in the initial positioning image and the centroid of the second region of the abnormal color block area in the re-examination positioning image, obtain the centroid offset direction and determine the partition category of the adjacent parts in the centroid offset direction.

[0113] It's important to note that the centroid of a region is the center of mass of the abnormal color patch area. For a two-dimensional region in an image, the centroid is the average of the coordinates of all pixels within that region. If the lesion spreads uniformly, the centroids of the first and second regions should be roughly at the same position. However, if the lesion spreads in a concentrated direction, the centroid of the second region will shift relative to the centroid of the first region, and the direction of this shift reflects the main direction of the lesion's spread.

[0114] Understandably, the analysis module calculates the average coordinates of all pixels within the abnormal color patch region in the initial localization image to obtain the centroid of the first region, and calculates the average coordinates of all pixels within the abnormal color patch region in the re-examination localization image to obtain the centroid of the second region. Then, based on the direction of the line connecting the centroids of the first and second regions, the direction of centroid offset is obtained, and the region partitioning map is consulted to determine the partition category of adjacent regions along the centroid offset direction. For example, if the abnormal color patch region is located in the palm partition, the coordinates of the centroid of the first region are (100, 150), and the coordinates of the centroid of the second region are (95, 145), then the direction of centroid offset is upward and to the left. The analysis module consults the region partitioning map and finds that the upper left adjacent region of the palm partition is the finger partition, which is classified as a high-sensitivity partition, indicating that the lesion is spreading towards the high-sensitivity partition.

[0115] Step C322: When the partition category is determined to be a high-sensitivity partition, the point diffusion amount is determined based on the difference between the second area and the first area and the third area of ​​the positioning acquisition frame.

[0116] Understandably, the spread of lesions to highly sensitive areas has a significant impact on patients. The analysis module calculates the difference between the second and first areas, uses the third area of ​​the localization acquisition frame as a normalization benchmark, and calculates the ratio of the difference between the second and first areas to the third area as the point diffusion amount. A positive value indicates diffusion, and a negative value indicates regression. For example, if the first area is 200 pixels, the second area is 280 pixels, the difference between the second and first areas is 80 pixels, and the third area is 2000 pixels, then the point diffusion amount is 80 / 2000 = 0.04.

[0117] Step C323: When the partition category is determined to be a low-sensitivity partition, the point diffusion amount is determined based on the area difference between the second area and the first area, the third area, and the centroid offset distance between the centroid of the first area and the centroid of the second area.

[0118] It should be noted that the clinical urgency of lesion spread to low-sensitivity areas is relatively low. The analysis module calculates the area change ratio by dividing the difference between the second and first areas by the third area, calculates the centroid offset distance between the centroids of the first and second regions using the Euclidean distance, and then calculates the offset ratio by dividing the centroid offset distance by the side length of the positioning acquisition frame. The difference between the area change ratio and the product of the offset ratio and the adjustment factor is taken as the point diffusion amount.

[0119] For example, the first area is 200 pixels, the second area is 280 pixels, the area difference between the second and first areas is 80 pixels, and the third area is 2000 pixels. The area change ratio is 80 divided by 2000, which equals 0.04. The coordinates of the centroid of the first area are (100, 150), the coordinates of the centroid of the second area are (110, 158), the centroid offset distance between the centroids of the first and second areas is 12.8 pixels, the side length of the positioning acquisition frame is 45 pixels, the offset ratio is 12.8 / 45≈0.28, and the adjustment factor is 0.1. Then the point diffusion amount is 0.04 - (0.28 × 0.1) = 0.012. Compared to the point diffusion amount of 0.04 when diffusing to high-sensitivity areas, the point diffusion amount when diffusing to low-sensitivity areas has been reduced, reflecting the difference in clinical significance between the two diffusion directions.

[0120] Preferably, steps C321 to C323 determine the diffusion direction by calculating the positional changes of the centroids of the first and second regions, and use different calculation methods according to the zoning categories of adjacent parts in the centroid offset direction, so that the diffusion amount at the point not only reflects the area change, but also reflects the clinical significance of the diffusion direction.

[0121] Preferably, step C3 eliminates the error caused by the difference in shooting distance by using the third area of ​​the positioning acquisition frame as a normalization benchmark, and the quantitative results can reflect the clinical significance of lesion changes by combining the diffusion direction for differential calculation.

[0122] Ideally, the analysis module guides patients to re-examine and photograph the same location by generating a positioning acquisition frame, which solves the problem of inconsistent positions when patients take multiple photos. Furthermore, it enables the accurate quantification of the spread or regression of lesions through normalization and direction-sensitive calculation methods.

[0123] The decision module determines the follow-up method based on the diffusion rate of the site and the location of the reaction zone in the site zoning map.

[0124] It's important to note that the analysis module has calculated the diffusion rate at the lesion site. A positive diffusion rate indicates that the lesion is spreading, while a negative value indicates that the lesion is regressing. However, even with diffusion, the clinical significance differs depending on whether it occurs in a highly sensitive area or a low-sensitivity area. Reactions in highly sensitive areas directly impact a patient's daily functions; for example, a reaction in the finger area may affect grasping, requiring immediate return to the hospital for a doctor's evaluation. Reactions in low-sensitivity areas have a relatively smaller impact on function, and monitoring can be strengthened by shortening the sampling interval; a return to the hospital is not necessary at this time. If the lesion is regressing, the sampling interval can be adjusted according to the degree of regression. Significant regression can extend the interval, while insignificant regression can maintain the original interval for continued observation.

[0125] Specifically, the execution steps of the decision-making module include steps D1 to D3: Step D1: If the diffusion rate of the detection point is positive and the zoning category of the reaction zone is a high-sensitivity zone, the follow-up method will be determined as a return to the hospital for re-examination.

[0126] Understandably, a positive value for the diffusion rate indicates that the lesion is spreading, and a high-sensitivity zone classification in the reaction zone indicates that the lesion is located in a functionally important area. The simultaneous fulfillment of both conditions means that the lesion in the functionally important area is worsening. The decision module then determines the follow-up method as a return to the hospital for re-examination, prompting the patient to go to the hospital as soon as possible for in-person evaluation and treatment by a doctor.

[0127] Step D2: When the point diffusion is positive and the partition category is low sensitivity partition, the follow-up method is determined to shorten the collection interval.

[0128] Understandably, a positive value for the diffusion rate indicates that the lesion is spreading, but a low-sensitivity zoning category suggests that the lesion is located in an area with less functional impact. In this case, although the lesion is developing, its urgency is not as high as in high-sensitivity zoning. The decision module determines the follow-up method to shorten the collection interval, for example, from once every three days to once a day, increasing the monitoring frequency to detect changes in the condition in a timely manner.

[0129] Step D3: When the diffusion rate of a point is negative, if the absolute value of the diffusion rate is greater than the ratio of the first area to the third area, the follow-up method is to extend the collection interval; if the absolute value of the diffusion rate is less than or equal to the ratio of the first area to the third area, the follow-up method is to maintain the collection interval.

[0130] It should be noted that a negative diffusion value indicates that the lesion is receding, but the degree of receding varies, and the follow-up methods should also differ accordingly. The ratio of the first area to the third area reflects the proportion of the abnormal color patch area in the positioning acquisition frame at the time of the initial acquisition, and can be used as a reference benchmark for judging the degree of receding.

[0131] Understandably, if the absolute value of the lesion diffusion is greater than the ratio of the first area to the third area, it indicates a significant reduction in lesion size and good recovery. In this case, the decision module will determine the follow-up method to extend the collection interval, for example, changing from once a day to once every three days, to reduce the patient's collection burden. If the absolute value of the lesion diffusion is less than or equal to the ratio of the first area to the third area, it indicates that although the lesion is receding, the reduction is not significant, and continued observation is needed. In this case, the decision module will determine the follow-up method to maintain the collection interval and keep the current monitoring frequency.

[0132] The modules in the aforementioned dynamic monitoring and follow-up management system for adverse reactions after chemotherapy can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0133] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a dynamic monitoring and follow-up management system for adverse reactions after chemotherapy. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0134] Those skilled in the art will understand that Figure 4The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0135] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0136] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0137] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0138] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0139] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0140] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0141] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dynamic monitoring and follow-up management system for adverse reactions after chemotherapy, characterized in that, include: The acquisition module is used to determine the symptoms to be monitored and the corresponding target sites based on the patient's chemotherapy regimen, and to receive images of the target sites from the patient's terminal. The identification module is used to register the part partition map corresponding to the target part and the target part image, determine the part partition with abnormal color block area as the reaction partition, and mark the color block positioning point according to the position of the abnormal color block area; The analysis module is used to generate a positioning acquisition frame according to the part of the location to which the color block positioning point belongs, receive the follow-up image of the patient terminal taken according to the positioning acquisition frame, crop the target part image and the follow-up part image based on the positioning acquisition frame, and determine the point diffusion amount. The decision module is used to determine the follow-up method based on the diffusion amount of the point and the zoning position of the reaction zone in the site zoning map.

2. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy according to claim 1, characterized in that: The process of determining the symptoms to be monitored and the corresponding target sites based on the patient's chemotherapy regimen includes: Multiple candidate sites are identified based on the chemotherapy drug information in the chemotherapy regimen, and the site incidence rate of each candidate site is obtained; When the number of chemotherapy cycles in the chemotherapy regimen is determined to be the first cycle, the candidate site with the highest incidence rate is selected as the target site. When the number of chemotherapy cycles is determined to be a non-first cycle, candidate sites with reaction zones in previous chemotherapy cycles are retrieved as target sites.

3. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy according to claim 1, characterized in that: The registration of the region partition map corresponding to the target region and the target region image includes: Color block recognition is performed on the target area image, and abnormal color block areas in the target area image are marked as registration exclusion areas; After excluding the registration exclusion region in the target region image, the region contour of the target region is identified, and multiple contour feature points on the region contour are extracted; Obtain multiple boundary feature points on the outer boundary of the partition of the part partition map, calculate the contour feature spacing between adjacent contour feature points, and calculate the boundary feature spacing between adjacent boundary feature points. The region partition map is scaled based on the contour feature spacing and boundary feature spacing. The boundary feature points in the scaled region partition map are then aligned and superimposed with the contour feature points to complete the registration.

4. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy according to claim 3, characterized in that: The extraction of multiple contour feature points on the contour of the region includes: Calculate the average contour curvature of the part and the contour curvature of each contour point, and take the contour points whose contour curvature is less than the average contour curvature as candidate concavity positions. Obtain the boundary positions between adjacent part partitions in the part partitioning map, and count the number of the boundary positions as the number of anatomical depressions; Based on the number of candidate depression locations and the number of anatomical depressions, contour feature points are determined.

5. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy according to claim 4, characterized in that: The step of determining contour feature points based on the number of candidate depression locations and the number of anatomical depressions includes: When the number of candidate depression locations is equal to the number of anatomical depressions, the candidate depression locations are used as contour feature points. When the number of candidate depression locations is greater than the number of dissected depressions, calculate the positional distance between each candidate depression location and the boundary location of each partition, the candidate distance between adjacent candidate depression locations, and the average candidate distance. If the position spacing is determined to be the minimum value among the position spacings between each concave candidate position and the boundary position of the same partition, and the difference between the candidate spacing between the concave candidate position and the adjacent concave candidate position and the average value of the candidate spacing is less than the average value of the candidate spacing, then the concave candidate position is taken as a contour feature point. If the number of candidate depression locations is less than the number of anatomical depressions, a re-acquisition prompt is sent to the patient terminal, and the re-acquisition prompt includes a location expansion prompt.

6. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy as described in claim 1, characterized in that: The area partitioned to identify regions with abnormal color patches is called a reaction partition, which includes: The corresponding region of the registration exclusion region in the registered target region image is taken as the first region, and the region in each region partition other than the first region is taken as the second region. If the partition category of the part where the first region is located is determined to be a high-sensitivity partition, or if the partition category is a low-sensitivity partition and there is a marked abnormal color block area in the high-sensitivity partition, then the first region is marked as an abnormal color block area. When it is determined that the partition category is a low-sensitivity partition and there are no marked abnormal color block areas in the high-sensitivity partition, when it is determined that the first color difference between the first area and the second area is greater than the partition color difference between the second area of ​​the high-sensitivity partition and the low-sensitivity partition, the first area is marked as an abnormal color block area. The areas containing the abnormal color blocks are divided into reaction zones.

7. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy as described in claim 6, characterized in that: The step of marking color block positioning points according to the location of the abnormal color block area includes: When the number of abnormal color block regions is determined to be a single one, the center of the abnormal color block region is taken as the color block positioning point; When the number of abnormal color block regions is determined to be multiple, the center of the abnormal color block region with the largest area in the abnormal color block region located in the high-sensitivity partition is taken as the color block positioning point. If all abnormal color block regions are located in the low-sensitivity partition, the center of the abnormal color block region with the largest area is taken as the color block positioning point. When it is determined that the abnormal color block area spans adjacent part partitions, the part partition with the largest area of ​​the abnormal color block area within each part partition is taken as the reaction partition, and the center of the part of the abnormal color block area within the reaction partition is taken as the color block positioning point.

8. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy as described in claim 1, characterized in that: The step of generating a positioning acquisition frame based on the location of the color block positioning point includes: The size of the acquisition frame is determined based on the area size of the abnormal color block region corresponding to the color block positioning point and the partition boundary of the part to which the color block positioning point belongs. A positioning acquisition frame is generated with the color block positioning point as the center and based on the acquisition frame size; The process of cropping the target area image and the review area image according to the positioning acquisition frame, and determining the point diffusion, includes: The initial positioning image is obtained by cropping the target area image according to the positioning acquisition frame, and the re-examination positioning image is obtained by cropping the re-examination area image according to the positioning acquisition frame. Calculate the first area of ​​the abnormal color block region in the initial positioning image and the second area of ​​the abnormal color block region in the re-examination positioning image, and determine the point diffusion amount based on the second area and the first area.

9. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy as described in claim 8, characterized in that: The step of determining the point diffusion amount based on the second area and the first area includes: Based on the centroid of the first region of the abnormal color block area in the initial positioning image and the centroid of the second region of the abnormal color block area in the re-examination positioning image, the centroid offset direction is obtained and the partition category of adjacent parts in the centroid offset direction is determined. When the partition category is determined to be a high-sensitivity partition, the point diffusion amount is determined based on the difference between the second area and the first area and the third area of ​​the positioning acquisition frame; When the partition category is determined to be a low-sensitivity partition, the point diffusion amount is determined based on the area difference between the second area and the first area, the third area, and the centroid offset distance between the centroid of the first region and the centroid of the second region.

10. The dynamic monitoring and follow-up management system for adverse reactions after chemotherapy as described in claim 1, characterized in that: The step of determining the follow-up method based on the diffusion amount at the site and the zoning location of the reaction zone in the site zoning map includes: When the diffusion amount at the point is determined to be positive and the reaction zone is classified as a highly sensitive zone, the follow-up method will be determined to be a return to the hospital for re-examination. When the point diffusion is determined to be positive and the partition category is low sensitivity partition, the follow-up method is to shorten the collection interval; When the diffusion amount at a given point is negative, and the absolute value of the diffusion amount is greater than the ratio of the first area to the third area, the follow-up method is determined to be extending the collection interval. When the absolute value of the diffusion amount at a given point is less than or equal to the ratio of the first area to the third area, the follow-up method is determined to be maintaining the collection interval.

Images