Method and system for predicting quality of life after breast cancer surgery based on dynamic changes of bone density

Abstract

The present invention discloses a method and system for predicting the quality of life after breast cancer surgery based on dynamic changes in bone density, and relates to the field of medical information analysis technology. The method and system include collecting sample data; extracting bone density data features to obtain dynamic bone density features; calculating the psychological threshold of each sample; calculating the time turning point and dividing the time period according to the time turning point; establishing a turning point prediction model for different time turning points; establishing a quality of life prediction model for different time periods, and predicting the test sample based on the time period in which the time point to be predicted is located using the prediction model of the target time period. The present invention divides the patient's postoperative time into time periods by combining the dynamic changes in bone density with the patient's age, treatment method, and psychological resilience score, establishes different quality of life prediction models for different time periods, and utilizes the different psychological fluctuations of patients in different time periods to improve the accuracy of the prediction results.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information analysis, and in particular to a method and system for predicting quality of life after breast cancer surgery based on dynamic changes in bone density. Background Art

[0002] The prediction of quality of life after breast cancer surgery is of great significance. It can help doctors develop personalized rehabilitation plans, prepare patients mentally, and optimize treatment decisions. Its influencing factors include physiological surgical complications, tumor recurrence and metastasis, endocrine changes, psychological anxiety and depression, social support, social economic status, and changes in work and life roles.

[0003] Prediction methods include clinical evaluation, questionnaires, and statistical models. For example, patent publication number CN117292838A describes a method and device for predicting the risk of psychological distress in patients undergoing chemotherapy after surgery for breast cancer. The method comprises: obtaining an initial sample data set of patients undergoing chemotherapy after surgery for breast cancer, the initial sample data set including the patient's psychological distress data and initial influencing factor data related to the patient's degree of psychological distress; screening the initial influencing factor data based on the correlation between the initial influencing factor data and the patient's psychological distress data, determining the main influencing factor data and its corresponding main sample data set; training the constructed psychological distress risk prediction model based on the main sample data set to obtain a fully trained psychological distress risk prediction model, and predicting the patient's psychological distress risk based on the fully trained psychological distress risk prediction model. This achieves a high-precision and rapid assessment of the patient's psychological distress risk.

[0004] Since breast cancer treatments such as surgery, chemotherapy, and endocrine therapy often lead to a decrease in estrogen levels in patients, increased osteoclast activity, and suppressed osteoblast activity, which in turn leads to a decrease in bone density and an increased risk of osteoporosis and fractures, changes in bone density will have a significant impact on the patient's quality of life. The accuracy of predicting the patient's postoperative quality of life based on the dynamic changes in bone density is relatively high. However, in real life, the patient's postoperative quality of life will produce different psychological supplementary effects in different time periods. The above method and other methods based on the same principle cannot present the quality of life prediction results at different time points based on the patient's psychological changes in different time periods, and the prediction results are not accurate enough. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for predicting the quality of life after breast cancer surgery based on dynamic changes in bone density, so as to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density, comprising:

[0007] Collect sample data: Obtain the age, treatment method, psychological resilience score, bone density data and quality of life score of several patients at different time points as historical samples; obtain the age, treatment method, psychological resilience score and bone density data of the patients to be predicted at different time points as test samples, and obtain the time point for prediction;

[0008] Data preprocessing: Data preprocessing is performed through data deletion and supplementation methods, and data coding of treatment methods is performed to improve historical sample data;

[0009] Bone density data feature extraction: Bone density data feature extraction is performed on historical samples and test samples using a time series feature extraction method to obtain dynamic bone density features;

[0010] Psychological threshold calculation: According to the psychological threshold calculation method, the psychological threshold of each sample in the historical sample is calculated through the psychological resilience in the historical sample, and the change range of the quality of life is reflected through the psychological threshold;

[0011] Sample data processing: The time turning point of the quality of life score of each sample in the historical sample is calculated by the psychological threshold and time turning point calculation method. The patient's postoperative time is divided into time periods according to the time turning point. Different postoperative time periods are divided according to the different psychological thresholds of each patient to improve the accuracy of time period division;

[0012] Predicting turning points: Based on the dynamic characteristics of bone density and the turning point prediction method, a turning point prediction model for different time turning points is established in sequence, the time turning point of the patient to be predicted is calculated, and the postoperative time of the patient to be predicted is divided into time periods;

[0013] Predicting quality of life: Based on the dynamic characteristics of bone density and the quality of life prediction method, a quality of life prediction model for different time periods is established through historical samples. According to the time period in which the time point to be predicted is located, the prediction model of the target time period is used to predict the test samples and output the quality of life prediction result for the target time point.

[0014] Preferably, the psychological threshold calculation method includes:

[0015] Based on the age and psychological resilience scores of patients in the historical sample, the psychological threshold of the target patient is calculated as follows:

[0016]

[0017] in represents the psychological threshold of the target patient, represents the psychological resilience score of the target patient, Indicates the age of the target patient. The psychological resilience score is combined with the patient's age through logarithmic operation, and the number of values ​​is increased to improve the accuracy of the expression of the patient's psychological state.

[0018] Preferably, the time turning point calculation method includes:

[0019] The same number of quality of life scores were selected at the time points before and after each quality of life score, and variance calculation was performed. The calculation result was used as the relative variance of the target quality of life score.

[0020] The relative variance of each quality of life score was compared with the psychological threshold, and the time nodes corresponding to multiple consecutive quality of life scores with relative variances greater than or less than the psychological threshold were divided into the same time period. Starting from the second time period, the first time node of each time period was used as the time turning point. Based on the changing trend of the relative variance and the different psychological thresholds of each patient, different postoperative time periods were divided to improve the accuracy of time period division.

[0021] Preferably, the turning point prediction method includes:

[0022] The patient's age, treatment method, psychological resilience score, bone density dynamic characteristics, and multiple time turning points in the statistical historical sample were analyzed. The bone density dynamic characteristics included the average rate of change of bone density and the standard deviation of the rate of change of bone density;

[0023] A turning point prediction model is established for each time turning point, specifically:

[0024] ,

[0025]

[0026] in represents the predicted moment of the first turning point after time, represents the predicted moment of the t-th turning point, and t>1, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. and Represent the regression constants in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and Represent the regression coefficients of age in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and They represent the regression coefficients of treatment methods in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and are the regression coefficients of the psychological resilience scores in the prediction model of the first turning point and the prediction model of the t-th turning point, respectively. and They represent the regression coefficients of the average change rate of bone density in the prediction model of the first time turning point and the prediction model of the t-th time turning point, respectively. and They represent the regression coefficients of the standard deviation of the bone density change rate in the first time turning point prediction model and the t-th time turning point prediction model, respectively. represents the predicted moment of the r-th time turning point, Represents the regression coefficient at the predicted moment of the r-th turning point;

[0027] By minimizing the loss function, each turning point prediction model is trained in chronological order to obtain the regression coefficient and regression constant;

[0028] The age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample are sequentially brought into each turning point prediction model to obtain the predicted moment of each time turning point. Each turning point prediction model is based on the previous time turning point and utilizes the correlation between time turning points to improve the prediction accuracy of time turning points.

[0029] Preferably, the quality of life prediction method includes:

[0030] Splitting the historical samples according to the divided time periods, the dynamic characteristics of bone density include the average change rate of bone density and the standard deviation of the change rate of bone density;

[0031] Establish a quality of life prediction model for different time periods, specifically:

[0032] ,

[0033]

[0034] in Indicates the position coefficient of the time point to be predicted in the y-th time period. Indicates the time point that needs to be predicted. Indicates the previous turning point of the y-th time period of the time point to be predicted. Indicates the next turning point in the y-th time period of the time point to be predicted. represents the predicted results of quality of life at the predicted time point, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. represents the regression constant in the y-th time period, 、 、 、 and represent the regression coefficients of age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density in the yth time period;

[0035] By minimizing the loss function, the quality of life prediction model of each time period is trained in chronological order to obtain 、 、 、 、 and The value of

[0036] Determine the time period in which the time point to be predicted is located, and sequentially bring the age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample into the quality of life prediction model corresponding to the target time period to obtain the quality of life prediction result at the predicted time point. Use the differences in the psychological fluctuations of patients in different time periods to improve the accuracy of the prediction results.

[0037] Preferably, the time series feature extraction method includes:

[0038] Calculate the bone density change rate between each two time nodes in the historical sample and the test sample: select two consecutive time nodes, calculate the difference in bone density data between the two time nodes, and then calculate the ratio between the difference and the previous bone density data to obtain the bone density change rate between the two target time nodes;

[0039] The average bone density change rate and the standard deviation of the bone density change rate were calculated from the obtained multiple bone density change rates as the dynamic characteristics of bone density.

[0040] Preferably, the data deletion and supplementation method includes:

[0041] In the historical sample, samples with missing age, treatment method or psychological resilience score were deleted;

[0042] For samples with missing bone density data: calculate the geometric ratio of the bone density at the previous moment to the bone density at the next moment to fill the gap;

[0043] For samples with missing quality of life scores, the geometric ratio of the quality of life scores at the previous moment and the quality of life scores at the next moment was calculated to fill the gap.

[0044] Preferably, the minimization loss function uses a mean square error function to estimate the parameters of the model, specifically:

[0045]

[0046] in represents the mean square error between the predicted value and the actual value, represents the total number of samples used to train the model, represents the actual value of d samples, Represents the model's predicted value for d samples;

[0047] By adjusting the coefficients in the model, Minimize the optimal model coefficient.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The dynamic characteristics of bone density are extracted through the dynamic changes of bone density. Then, the patient's postoperative time is divided into time periods based on the patient's age, treatment method and psychological resilience score. Different time periods represent different psychological fluctuations of the patient. Different quality of life prediction models are established in different time periods. The quality of life prediction results at a specific time point are calculated based on the prediction model. The differences in the psychological fluctuations of patients in different time periods are used to improve the accuracy of the prediction results.

[0050] At the same time, when dividing the time periods, the psychological resilience score is first combined with age to calculate the patient's psychological threshold, and the sensitivity of the psychological threshold is improved to improve the accuracy of the expression of the patient's psychological state. Then, when dividing the time periods, different postoperative time periods are divided according to the different psychological thresholds of each patient to improve the accuracy of the time period division. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the process of the quality of life prediction method of the present invention;

[0052] Figure 2 Schematic diagram of the process of extracting time series features in the present invention;

[0053] Figure 3 Schematic diagram of the flow of the time turning point calculation method of the present invention;

[0054] Figure 4 Schematic diagram of the turning point prediction method of the present invention;

[0055] Figure 5 Schematic diagram of the process of the quality of life prediction method of the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] In this application, for ease of understanding, the method steps used do not need to be executed in the order of the steps in this embodiment during actual operation. In other embodiments, these steps may be performed simultaneously or in a different order.

[0058] Breast cancer treatments often lead to decreased bone density in patients, increasing the risk of osteoporosis and fractures. Therefore, changes in bone density can have a significant impact on patients' quality of life. Predicting patients' postoperative quality of life based on dynamic changes in bone density is highly accurate. In real life, patients' postoperative quality of life will produce different psychological supplementary effects at different time periods. Predicting quality of life based on patients' psychological changes at different time periods can improve the accuracy of the prediction results.

[0059] like Figure 1-Figure 5 As shown, the present invention provides a technical solution: a method and system for predicting quality of life after breast cancer surgery based on dynamic changes in bone density, comprising:

[0060] Collect sample data: Obtain the age, treatment method, psychological resilience score, bone density data and quality of life score of several patients at different time points as historical samples; obtain the age, treatment method, psychological resilience score and bone density data of the patients to be predicted at different time points as test samples, and obtain the time point for prediction;

[0061] Data preprocessing: Data preprocessing is performed through data deletion and supplementation methods, and data coding of treatment methods is performed to improve historical sample data;

[0062] Bone density data feature extraction: Bone density data feature extraction is performed on historical samples and test samples using a time series feature extraction method to obtain dynamic bone density features;

[0063] Psychological threshold calculation: According to the psychological threshold calculation method, the psychological threshold of each sample in the historical sample is calculated through the psychological resilience in the historical sample, and the change range of the quality of life is reflected through the psychological threshold;

[0064] Sample data processing: The time turning point of the quality of life score of each sample in the historical sample is calculated by the psychological threshold and time turning point calculation method. The patient's postoperative time is divided into time periods according to the time turning point. Different postoperative time periods are divided according to the different psychological thresholds of each patient to improve the accuracy of time period division;

[0065] Predicting turning points: Based on the dynamic characteristics of bone density and the turning point prediction method, a turning point prediction model for different time turning points is established in sequence, the time turning point of the patient to be predicted is calculated, and the postoperative time of the patient to be predicted is divided into time periods;

[0066] Predicting quality of life: Based on the dynamic characteristics of bone density and the quality of life prediction method, a quality of life prediction model for different time periods is established through historical samples. According to the time period in which the time point to be predicted is located, the prediction model of the target time period is used to predict the test samples and output the quality of life prediction result for the target time point.

[0067] It should be noted that the patient's psychological resilience score can be obtained through the psychological resilience scale (CD-RISC), and the patient's quality of life score can be assessed through the EORTC QLQ-C30 scale.

[0068] Data deletion and supplementation methods include:

[0069] In the historical sample, samples with missing age, treatment method or psychological resilience score were deleted;

[0070] For samples with missing bone density data: calculate the geometric ratio of the bone density at the previous moment to the bone density at the next moment to fill the gap;

[0071] For samples with missing quality of life scores: calculate the geometric ratio of the quality of life scores at the previous moment and the quality of life scores at the next moment to complete the missing score;

[0072] The geometric value is calculated as follows: according to the number of missing items, the difference between the previous and next values ​​is divided into equal proportions, and then added to the missing items in sequence so that the difference between the two adjacent items is the same;

[0073] Treatments were numbered sequentially.

[0074] It should be noted that for ease of understanding, simulated data is used as follows:

[0075] For a patient in a historical sample, the data records are:

[0076] Age: 38

[0077] The treatment is: breast-conserving surgery + endocrine therapy;

[0078] Mental toughness score: 68;

[0079] Bone density data (postoperative time, unit is g / cm², the same unit below) are: first week: 1.26; second week 1.24; third week: missing; fourth week: 1.22...;

[0080] Quality of life score: Week 1: 80; Week 2: missing; Week 3: missing; Week 4: 77...

[0081] In this example, the quality methods are numbered according to the surgical method (breast-conserving and resection) and whether endocrine therapy is performed. Breast-conserving plus endocrine therapy is numbered 1, breast-conserving plus no endocrine therapy is numbered 2, resection plus endocrine therapy is numbered 3, and resection plus no endocrine therapy is numbered 4.

[0082] For bone density data: the geometric ratio values ​​of the data from the second and fourth weeks after surgery were used to supplement the data. Since there was only one missing value in the middle, the middle value was taken as 0.5×(1.24+1.22)=1.23, so the bone density data of the third week was 1.23;

[0083] For the quality of life score: the data from the first and fourth weeks after surgery were used to supplement the data from the second and third weeks. Since two items were missing, the missing data were increased proportionally by one-third, (80-77) ÷ 3 = 1, 80-1 = 79, 79-1 = 78. Therefore, the quality of life score for the second week was 79, and the quality of life score for the third week was 78.

[0084] like Figure 2 As shown in Figure 2, the time series feature extraction method includes:

[0085] Calculate the bone density change rate between each two time nodes in the historical samples and the test samples: select two consecutive time nodes, calculate the difference in bone density data between the two time nodes, and then calculate the ratio between the difference and the previous bone density data to obtain the bone density change rate between the two target time nodes, specifically:

[0086]

[0087] in represents the rate of change of bone density between time j and time j+1 in the i-th sample, represents the bone density data value at the jth moment in the i-th sample, represents the bone density data value at the j+1th moment in the i-th sample;

[0088] The average bone density change rate and the standard deviation of the bone density change rate are calculated by obtaining multiple bone density change rates as the dynamic characteristics of bone density, specifically:

[0089] ,

[0090]

[0091] in represents the average rate of change of bone density, Indicates the total number of time nodes, represents the rate of change of bone density between time j and time j+1 in the i-th sample, represents the standard deviation of the rate of change of bone density.

[0092] It should be noted that for ease of understanding, simulated data is used as follows:

[0093] For patients in the historical sample, the multiple bone density measurements were 1.20, 1.18, 1.16, and 1.14;

[0094] For this patient, the bone density change rate between the first and second time nodes was (1.18-1.20) ÷ 1.20 ≈ -0.0167. Using the same method, the bone density change rate between the second and third time nodes was approximately -0.0169, and the bone density change rate between the third and fourth time nodes was approximately -0.0172.

[0095] By calculating the average change rate of bone density, we found that (-0.0167-0.0169-0.0172) ÷ 3 ≈ -0.0169;

[0096] The standard deviation of the bone density change rate was approximately 0.0002.

[0097] The dynamic characteristics of bone density of all patients in the historical sample can be calculated in the same way.

[0098] Psychological threshold calculation methods include:

[0099] Based on the age and psychological resilience scores of patients in the historical sample, the psychological threshold of the target patient is calculated as follows:

[0100]

[0101] in represents the psychological threshold of the target patient, represents the psychological resilience score of the target patient, Indicates the age of the target patient. The psychological resilience score is combined with the patient's age through logarithmic operation, and the number of values ​​is increased to improve the accuracy of the expression of the patient's psychological state.

[0102] It should be noted that for ease of understanding, simulated data is used as follows:

[0103] For patients in the historical sample, whose psychological resilience score is 79 and whose age is 50, their psychological threshold is lg(79+1)+lg(50)≈3.6021.

[0104] By performing logarithmic operations on the psychological resilience score and the patient's age, the value accuracy is improved compared to the psychological resilience score, and it can more accurately reflect the patient's psychological state.

[0105] like Figure 3 As shown, the time turning point calculation method includes:

[0106] The same number of quality of life scores were selected at the time points before and after each quality of life score, and variance calculation was performed. The calculation result was used as the relative variance of the target quality of life score, specifically:

[0107]

[0108] in represents the relative variance of the target quality of life score, represents the total number of selected quality of life scores, Indicates the value of the nth data in the selected quality of life score, Indicates the selected The mean of the quality of life scores;

[0109] The relative variance of each quality of life score was compared with the psychological threshold, and the time nodes corresponding to multiple consecutive quality of life scores with relative variances greater than or less than the psychological threshold were divided into the same time period. Starting from the second time period, the first time node of each time period was used as the time turning point. Based on the changing trend of the relative variance and the different psychological thresholds of each patient, different postoperative time periods were divided to improve the accuracy of time period division.

[0110] It should be noted that for the convenience of calculation, the simulated data are as follows:

[0111] The quality of life scores of patients every day after surgery were: 78, 79, 78, 77, 78, 76, 70, 65, 60, 58, 57, 57, 56;

[0112] Calculate the relative variance of each number, and select a value before and after the data when calculating the relative variance, for a total of five data values:

[0113] Starting from the second day after surgery (two values ​​can be taken forward), the relative variances on the second day after surgery and thereafter are approximately 0.22, 0.67, 0.22, 0.67, 11.56, 20.22, 16.67, 8.67, 1.56, 0.22, and 0.22, respectively. Assuming that the patient's psychological threshold is 3.6021, the relative variances higher than 3.6021 appear from the sixth to the ninth day after surgery, and the relative variances lower than 3.6021 appear from the first to the third day and after the tenth day. It can be judged that in the first five days The quality of life score is relatively stable and is divided into the first time period. The sixth to ninth days are divided into the second time period, and the tenth day and later are divided into the third time period. The first time period indicates that the patient's quality of life is good and relatively stable. The second time period indicates that the sequelae of treatment gradually affect life, and the quality of life score drops significantly. The third time period indicates that the quality of life score reaches the bottom and is relatively stable (in actual situations, patients are affected by more external factors and may be divided into more stages, such as the recovery stage after adapting to life), and the turning points are on the sixth and tenth days respectively.

[0114] like Figure 4 As shown, the turning point prediction methods include:

[0115] The patient's age, treatment method, psychological resilience score, bone density dynamic characteristics, and multiple time turning points in the statistical historical sample were analyzed. The bone density dynamic characteristics included the average rate of change of bone density and the standard deviation of the rate of change of bone density;

[0116] A turning point prediction model is established for each time turning point, specifically:

[0117] ,

[0118]

[0119] in represents the predicted moment of the first turning point after time, represents the predicted moment of the t-th turning point, and t>1, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. and Represent the regression constants in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and Represent the regression coefficients of age in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and They represent the regression coefficients of treatment methods in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and are the regression coefficients of the psychological resilience scores in the prediction model of the first turning point and the prediction model of the t-th turning point, respectively. and They represent the regression coefficients of the average change rate of bone density in the prediction model of the first time turning point and the prediction model of the t-th time turning point, respectively. and They represent the regression coefficients of the standard deviation of the bone density change rate in the first time turning point prediction model and the t-th time turning point prediction model, respectively. represents the predicted moment of the r-th time turning point, Represents the regression coefficient at the predicted moment of the r-th turning point;

[0120] By minimizing the loss function, each turning point prediction model is trained in chronological order to obtain the regression coefficient and regression constant;

[0121] The age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample are sequentially brought into each turning point prediction model to obtain the predicted moment of each time turning point. Each turning point prediction model is based on the previous time turning point and utilizes the correlation between time turning points to improve the prediction accuracy of time turning points.

[0122] It should be noted that for ease of understanding, simulated data is used as shown in Table 1:

[0123] Table 1: Historical sample table

[0124]

[0125] The first time turning point prediction model is trained using the above data, and the regression coefficients and regression constants are obtained in turn. The training method is as follows:

[0126] In this embodiment, the mean square error function is used to estimate the parameters of the model, specifically:

[0127]

[0128] in represents the mean square error between the predicted value and the actual value, represents the total number of samples used to train the model, represents the actual value of d samples, Represents the model's predicted value for d samples;

[0129] By adjusting the coefficients in the model, Minimize, thus obtaining the optimal model coefficient;

[0130] After bringing in the above data, we can calculate it by the least square method. ≈2.02, ≈0.03, ≈0.40, ≈0.02, ≈-104.75, ≈-43.55, from which the prediction model of the first time turning point can be determined. Similarly, according to the data in Table 1, the prediction model of the second time turning point and the prediction model of the third time turning point can be obtained in sequence, which will not be repeated here.

[0131] Assuming that the age of the patient to be predicted is 38, the treatment method is 2, the psychological resilience score is 70, the average change rate of bone density is -0.0166, and the standard deviation of the bone density change rate is 0.0002, it is brought into the prediction model of the first time turning point to obtain the predicted moment of the first time turning point of the patient to be tested. ≈7.09, so by rounding it off we can predict that the first turning point in time is the seventh day.

[0132] Then the predicted time of the first time turning point is brought into the second time turning point prediction model to calculate the predicted moment of the second time turning point. The algorithm for predicting the third time turning point is the same and will not be repeated here. Each turning point prediction model is based on the previous time turning point and uses the correlation between time turning points to improve the prediction accuracy of the time turning point.

[0133] like Figure 5 As shown, quality of life prediction methods include:

[0134] Splitting the historical samples according to the divided time periods, the dynamic characteristics of bone density include the average change rate of bone density and the standard deviation of the change rate of bone density;

[0135] Establish a quality of life prediction model for different time periods, specifically:

[0136] ,

[0137]

[0138] in Indicates the position coefficient of the time point to be predicted in the y-th time period. Indicates the time point that needs to be predicted. Indicates the previous turning point of the y-th time period of the time point to be predicted. Indicates the next turning point in the y-th time period of the time point to be predicted. represents the predicted results of quality of life at the predicted time point, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. represents the regression constant in the y-th time period, 、 、 、 and represent the regression coefficients of age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density in the yth time period;

[0139] By minimizing the loss function, the quality of life prediction model of each time period is trained in chronological order to obtain 、 、 、 、 and The value of

[0140] Determine the time period in which the time point to be predicted is located, and sequentially bring the age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample into the quality of life prediction model corresponding to the target time period to obtain the quality of life prediction result at the predicted time point. Utilize the differences in the psychological fluctuations of patients in different time periods to improve the accuracy of the prediction results.

[0141] It should be noted that for ease of understanding, simulated data is used as shown in Table 2:

[0142] Table 2: Quality of life score sheet

[0143]

[0144] According to the data in Tables 1 and 2, the position coefficient of each sample at each measurement was calculated. For example, the first measurement of sample 1 was on the third day after surgery, and its first time period was from the first to the fifth day (before the first turning point), so the position coefficient was 0.5. The corresponding position moment algorithms for the corresponding measurement times of other samples were the same and will not be elaborated here.

[0145] Then, according to the mean square error function, the quality of life prediction model for each time period is trained using the data in Table 1 and Table 2 to obtain the parameters that are most suitable for the time period. The least squares method can be used for calculation. The calculation process is the same as that of the turning point prediction model and will not be described in detail here.

[0146] Assuming that the time point to be predicted is the third day, and the first time period with the patient to be predicted is from the first to the fifth day, the position coefficient of the patient to be predicted at this time node can be calculated to be 0.5. The information of the patient to be predicted can be brought into the trained quality of life prediction model of the first time period to obtain the prediction result.

[0147] By taking advantage of the differences in patients' psychological fluctuations in different time periods, prediction models with different coefficients are established to improve the accuracy of the prediction results.

[0148] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is limited by the accompanying embodiments and their equivalents.

Claims

1. A method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density, characterized by: include: Collect sample data: Obtain the age, treatment method, psychological resilience score, bone density data and quality of life score of several patients at different time points as historical samples; obtain the age, treatment method, psychological resilience score and bone density data of the patients to be predicted at different time points as test samples, and obtain the time point for prediction; Data preprocessing: Data preprocessing is performed through data deletion and supplementation methods, and data coding of treatment methods is performed to improve historical sample data; Bone density data feature extraction: Bone density data feature extraction is performed on historical samples and test samples using a time series feature extraction method to obtain dynamic bone density features; Psychological threshold calculation: According to the psychological threshold calculation method, the psychological threshold of each sample in the historical sample is calculated through the psychological resilience in the historical sample, and the change range of the quality of life is reflected through the psychological threshold; Sample data processing: The time turning point of the quality of life score of each sample in the historical sample is calculated by the psychological threshold and time turning point calculation method. The patient's postoperative time is divided into time periods according to the time turning point. Different postoperative time periods are divided according to the different psychological thresholds of each patient to improve the accuracy of time period division; Predicting turning points: Based on the dynamic characteristics of bone density and the turning point prediction method, a turning point prediction model for different time turning points is established in sequence, the time turning point of the patient to be predicted is calculated, and the postoperative time of the patient to be predicted is divided into time periods; Predicting quality of life: Based on the dynamic characteristics of bone density and the quality of life prediction method, a quality of life prediction model for different time periods is established through historical samples. According to the time period in which the time point to be predicted is located, the prediction model of the target time period is used to predict the test samples and output the quality of life prediction result for the target time point.

2. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The psychological threshold calculation method includes: Based on the age and psychological resilience scores of patients in the historical sample, the psychological threshold of the target patient is calculated as follows: ； in represents the psychological threshold of the target patient, represents the psychological resilience score of the target patient, Indicates the age of the target patient. The psychological resilience score is combined with the patient's age through logarithmic operation, and the number of values ​​is increased to improve the accuracy of the expression of the patient's psychological state.

3. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The time turning point calculation method includes: The same number of quality of life scores were selected at the time points before and after each quality of life score, and variance calculation was performed. The calculation result was used as the relative variance of the target quality of life score. The relative variance of each quality of life score was compared with the psychological threshold, and the time nodes corresponding to multiple consecutive quality of life scores with relative variances greater than or less than the psychological threshold were divided into the same time period. Starting from the second time period, the first time node of each time period was used as the time turning point. Based on the changing trend of the relative variance and the different psychological thresholds of each patient, different postoperative time periods were divided to improve the accuracy of time period division.

4. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The turning point prediction method comprises: The patient's age, treatment method, psychological resilience score, bone density dynamic characteristics, and multiple time turning points in the statistical historical sample were analyzed. The bone density dynamic characteristics included the average rate of change of bone density and the standard deviation of the rate of change of bone density; A turning point prediction model is established for each time turning point, specifically: , ； in represents the predicted moment of the first turning point after time, represents the predicted moment of the t-th turning point, and t>1, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. and Represent the regression constants in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and Represent the regression coefficients of age in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and They represent the regression coefficients of treatment methods in the first time turning point prediction model and the t-th time turning point prediction model, respectively. and are the regression coefficients of the psychological resilience scores in the prediction model of the first turning point and the prediction model of the t-th turning point, respectively. and They represent the regression coefficients of the average change rate of bone density in the prediction model of the first time turning point and the prediction model of the t-th time turning point, respectively. and They represent the regression coefficients of the standard deviation of the bone density change rate in the first time turning point prediction model and the t-th time turning point prediction model, respectively. represents the predicted moment of the r-th time turning point, Represents the regression coefficient at the predicted moment of the r-th turning point; By minimizing the loss function, each turning point prediction model is trained in chronological order to obtain the regression coefficient and regression constant; The age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample are sequentially brought into each turning point prediction model to obtain the predicted moment of each time turning point. Each turning point prediction model is based on the previous time turning point and utilizes the correlation between time turning points to improve the prediction accuracy of time turning points.

5. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The quality of life prediction method includes: Splitting the historical samples according to the divided time periods, the dynamic characteristics of bone density include the average change rate of bone density and the standard deviation of the change rate of bone density; Establish a quality of life prediction model for different time periods, specifically: ， ； in Indicates the position coefficient of the time point to be predicted in the y-th time period. Indicates the time point that needs to be predicted. Indicates the previous turning point of the y-th time period of the time point to be predicted. Indicates the next turning point in the y-th time period of the time point to be predicted. represents the predicted results of quality of life at the predicted time point, 、 、 、 and They represent the patient's age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density, respectively. represents the regression constant in the y-th time period, 、 、 、 and represent the regression coefficients of age, treatment method, psychological resilience score, average change rate of bone density and standard deviation of the change rate of bone density in the yth time period; By minimizing the loss function, the quality of life prediction model of each time period is trained in chronological order to obtain 、 、 、 、 and The value of Determine the time period in which the time point to be predicted is located, and sequentially bring the age, treatment method, psychological resilience score, and bone density dynamic characteristics of the patients in the test sample into the quality of life prediction model corresponding to the target time period to obtain the quality of life prediction result at the predicted time point. Utilize the differences in the psychological fluctuations of patients in different time periods to improve the accuracy of the prediction results.

6. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The time series feature extraction method includes: Calculate the bone density change rate between each two time nodes in the historical samples and the test samples: select two consecutive time nodes, calculate the difference in bone density data between the two time nodes, and then calculate the ratio between the difference and the previous bone density data to obtain the bone density change rate between the two target time nodes; The average bone density change rate and the standard deviation of the bone density change rate were calculated from the obtained multiple bone density change rates as the dynamic characteristics of bone density.

7. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 1, characterized in that: The data deletion and supplementation method includes: In the historical sample, samples with missing age, treatment method or psychological resilience score were deleted; For samples with missing bone density data: calculate the geometric ratio of the bone density at the previous moment to the bone density at the next moment to fill the gap; For samples with missing quality of life scores, the geometric ratio of the quality of life scores at the previous moment and the quality of life scores at the next moment was calculated to fill the gap.

8. The method for predicting quality of life after breast cancer surgery based on dynamic changes in bone density according to claim 4 or 5, characterized in that: The minimization loss function uses the mean square error function to estimate the parameters of the model, specifically: ； in represents the mean square error between the predicted value and the actual value, represents the total number of samples used to train the model, represents the actual value of d samples, Represents the model's predicted value for d samples; By adjusting the coefficients in the model, Minimize the optimal model coefficient.

9. A system for predicting quality of life after breast cancer surgery based on dynamic changes in bone density, characterized by: include: Data collection module: used to obtain the age, treatment method, psychological resilience score, bone density data and quality of life score of several patients at different time points as historical samples, obtain the age, treatment method, psychological resilience score and bone density data of the patient to be predicted at different time points as test samples, and obtain the time point required for prediction; Preliminary processing module: This module is used to preprocess data using data deletion and supplementation methods, encode treatment methods, and improve historical sample data. It also extracts bone density data features from historical samples and test samples using a time series feature extraction method to obtain dynamic bone density features. It also calculates the psychological threshold of each sample in the historical sample using the psychological resilience in the historical sample using a psychological threshold calculation method, and uses this threshold to reflect the magnitude of changes in quality of life. Central processing module: used to calculate the time turning point of the quality of life score of each sample in the historical samples through the psychological threshold and time turning point calculation method, divide the patient's postoperative time into time periods according to the time turning point, and divide the postoperative time periods into different time periods according to the different psychological thresholds of each patient, thereby improving the accuracy of time period division; based on the dynamic characteristics of bone density and the turning point prediction method, establish turning point prediction models for different time turning points in sequence, calculate the time turning point of the patient to be predicted, and divide the postoperative time of the patient to be predicted into time periods; based on the dynamic characteristics of bone density and the quality of life prediction method, establish quality of life prediction models for different time periods through historical samples; Data output module: It is used to predict the test samples according to the time period in which the time point to be predicted is located, through the prediction model of the target time period, and output the quality of life prediction result at the target time point.

Images