Data processing method and system for digestive tract tumors

Abstract

The present invention discloses a data processing method and system for digestive tract tumors, which relates to the field of medical information processing technology. By collecting the patient's key physiological parameters and tumor characteristics from multiple data sources, a tumor identification data set P of the patient is formed, and through standardized preprocessing and feature vector integration, the consistency and computability of the data are guaranteed. The disease progression prediction model constructed using a machine learning method takes the patient's feature vector X as input and can efficiently predict the patient's disease progression fluctuation probability index Progress. Furthermore, combined with the correction coefficient algorithm and the set correction fluctuation threshold Cthe, the recovery period content extracted from the patient's initial medical record can be dynamically corrected to obtain the corrected recovery period Tad. Through feedback and iterative optimization, the calculation formula of the corrected survival period is continuously verified and updated, ultimately achieving the effect of improving prediction accuracy and adapting to individual differences among patients.

Description

Technical Field

[0001] The present invention relates to the field of medical information processing technology, and in particular to a data processing method and system for digestive tract tumors. Background Art

[0002] In modern medicine, cancer diagnosis and treatment is a vast area of ​​research and application. Gastrointestinal cancers, including gastric, esophageal, and colorectal cancers, are common and fatal, posing a significant public health challenge worldwide. With advances in medical technology, progress has been made in the early diagnosis and treatment of gastrointestinal cancers.

[0003] Although doctors currently use traditional methods based on clinical experience in the assessment of the recovery period of gastrointestinal tumors, such as TNM staging, determination of tumor biomarkers, and imaging results, there are still many limitations. First, traditional recovery period prediction methods often rely on limited clinical indicators and cannot fully consider individual differences among patients, such as different age groups, different genetic backgrounds, and differences in treatment response. These factors may directly affect the rate of disease progression and treatment effectiveness. Therefore, when faced with complex cases, traditional methods are often unable to provide accurate prognostic assessments, leading to errors in treatment decisions. Furthermore, existing recovery period assessment methods are often relatively static and cannot be updated and adjusted in real time, and cannot make dynamic predictions based on changes in the patient's treatment process. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a data processing method and system for digestive tract tumors, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a data processing method for digestive tract tumors, comprising the following steps:

[0006] S1. Collect and obtain patient tumor marker data from the patient's electronic health record source, imaging examination record source, blood test record source, and genomic test record source to form the patient's tumor marker data set P;

[0007] S2. Perform data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrate the processed patient tumor marker data set P to obtain a feature vector X;

[0008] S3. Build a disease progression prediction model using a machine learning method, use the feature vector X as the input feature of the disease progression prediction model, and obtain the disease progression fluctuation probability index Progress predicted for the patient through training;

[0009] S4. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm, and comparing it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result. Based on the correction trigger result, the recovery period content extracted from the patient's initial medical record is corrected to obtain a corrected recovery period Tad;

[0010] S5. Perform feedback and iterative optimization based on the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

[0011] Preferably, said S1 includes S11 and S12;

[0012] S11. Collect individual difference-related parameters from clinical data sources, including extracting patient individual difference-related parameters from electronic health record sources, imaging examination record sources, blood test record sources, and genomic test record sources, including age A, sex S, weight W, height H, blood pressure BP1, gene mutation GM, and tumor size TS, to form a query collection data set CC = {A, S, W, H, BP, GM, TS};

[0013] Among them, age A, gender S, weight W and height H are obtained by entering the patient's unique information from the electronic health record source; tumor size TS is obtained by entering the patient's unique information from the imaging examination record source; blood pressure BP is obtained by entering the patient's unique information in the blood test record source to query the blood pressure measurement record; whether the gene is mutated GM is obtained by entering the patient's unique information in the genomic detection record source to obtain the gene sequence analysis results and mutation screening results.

[0014] Preferably, S12, integrating the query-collected data set CC={A, S, W, H, BP, GM, TS} to construct a patient tumor identification data set P;

[0015] Specifically, binary coding and categorical variable coding are performed on the sex S and whether the gene is mutated GM in the query data group CC to obtain the categorical variable sex Se and the binary gene mutation GMe;

[0016] Among them, the categorical variable sex Se includes classifying male as 1 and female as 2 to distinguish sex; the binary gene mutation GMe includes setting 0 as no gene mutation and 1 as the presence of gene mutation;

[0017] Substitute the categorical variable sex Se and the binary gene mutation GMe into the query data set CC = {A, S, W, H, BP, GM, TS} for replacement integration to obtain the integrated patient tumor identification data set P;

[0018] The patient tumor identification data set P = {A, Se, W, H, BP, GMe, TS}.

[0019] Preferably, said S2 includes S21;

[0020] S21. Perform data preprocessing on the patient's age A, weight W, height H, blood pressure BP, and tumor size TS in the tumor identification data set P, including standardization preprocessing, to obtain the preprocessed processed age Ac, processed weight Wc, processed height Hc, processed blood pressure BPc, and processed tumor size TSc, and substitute these data into the patient's tumor identification data set P for replacement, to obtain a feature vector X = {Ac, Se, Wc, Hc, BPc, GMe, TSc};

[0021] Among them, the standardization preprocessing eliminates the dimensional differences of the data by using feature scaling technology.

[0022] Preferably, said S3 includes S31 and S32;

[0023] S31. Construct a disease progression prediction model by using a machine learning method, and train the model using the feature vector X as an input feature to obtain a weighted linear combination z. Simultaneously, transform the weighted linear combination z using a Sigmoid function to obtain a predicted disease progression fluctuation probability index Progress for the patient.

[0024] The disease progression fluctuation probability index Progress is obtained by the following calculation formula:

[0025] z=(a1*Ac+a2*Se+a3*Wc+a4*Hc+a5*BPc+a6*GMe+a7*TSc+β);

[0026]

[0027] Where z represents a weighted linear combination, e represents a Sigmoid function, which is used to map the disease progression fluctuation probability index Progress to the range of 0 to 1, a1, a2, a3, a4, a5, a6, and a7 represent the weight coefficients of treatment age Ac, categorical variable sex Se, treatment weight Wc, treatment height Hc, treatment blood pressure BPc, binary gene mutation GMe, and treatment tumor size TSc, respectively, β represents a bias term, and a1+a2+a3+a4+a5+a6+a7=1. The specific value is set by the user.

[0028] Preferably, S32, using a supervised learning algorithm to find the optimal weight coefficient and bias term β for the disease progression prediction model, including using the cross entropy loss function to measure the difference between the prediction and the actual value, obtaining the total loss of the cross entropy loss function, and then using the gradient descent algorithm to gradually adjust the weight coefficient and bias term β to minimize the total loss, so as to iteratively optimize the constructed disease progression prediction model.

[0029] Preferably, said S4 includes S41 and S42;

[0030] S41. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm;

[0031] The correction coefficient Cm is obtained by the following calculation formula:

[0032]

[0033] Where B represents the exponential power, γ represents the influence coefficient of age A on the correction coefficient Cm, c1, c2 and c3 represent the influence coefficients on the correction coefficient Cm, c4 represents the nonlinear influence coefficient of weight W on the correction coefficient Cm, exp represents the exponential function, SZ represents the body mass index, and is specifically expressed by Calculation formula is obtained, W represents weight, H represents height.

[0034] Preferably, S42, the correction coefficient Cm is compared with a preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and the recovery period content extracted from the patient's initial medical record is corrected according to the correction trigger result to obtain the initial recovery period Tinit, and the corrected recovery period Tad is obtained by combining the correction coefficient Cm;

[0035] The correction trigger result is obtained by the following comparison method:

[0036] When the correction coefficient Cm is less than the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as a non-trigger result;

[0037] When the correction coefficient Cm ≥ the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as the trigger result;

[0038] The modified recovery period Tad is obtained by the calculation formula Tad=Tinit*(1+Cm).

[0039] Preferably, the S5 includes S51;

[0040] S51. Feedback and iterative optimization are performed based on the revised recovery period Tad, including verification and updating of the revised recovery period Tad. Verification includes comparing the revised recovery period Tad with the diagnostic recovery period data provided by the patient's attending physician to check the accuracy of the revised recovery period Tad and recording the recovery period difference ΔTad. When the recovery period difference ΔTad is less than 5%, the revised recovery period Tad is not optimized or updated. When the recovery period difference ΔTad is greater than or equal to 5%, the revised recovery period Tad is optimized and updated.

[0041] Among them, the optimization update includes using the recovery period difference △Tad to proportionally adjust the influence coefficient of the correction coefficient Cm and the weight coefficient of the disease progression prediction model.

[0042] The data processing system for digestive tract tumors includes a data acquisition module, a data preprocessing module, a modeling and analysis module, a correction and evaluation module, and an iterative optimization module;

[0043] The data acquisition module collects data from the patient's electronic health record source, imaging examination record source, blood test record source and genomic test record source to form the patient's tumor identification data set P;

[0044] The data preprocessing module performs data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrates the processed patient tumor marker data set P to obtain a feature vector X;

[0045] The modeling and analysis module constructs a disease progression prediction model by using a machine learning method, and uses the feature vector X as the input feature of the disease progression prediction model, and obtains the disease progression fluctuation probability index Progress predicted for the patient through training;

[0046] The correction evaluation module substitutes the acquired disease progression fluctuation probability index Progress into the correction coefficient algorithm formula to perform calculations to obtain the correction coefficient Cm, and compares it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and corrects the recovery period content extracted from the patient's initial medical record according to the correction trigger result to obtain a corrected recovery period Tad.

[0047] The iterative optimization module performs feedback and iterative optimization according to the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

[0048] The present invention provides a data processing method and system for digestive tract tumors, which has the following beneficial effects:

[0049] (1) Through steps S1 to S5, accurate prediction and dynamic adjustment of the patient's disease recovery progress are achieved. The patient's key physiological parameters and tumor characteristics are collected from multiple data sources to form the patient's tumor identification data set P. The consistency and computability of the data are ensured through standardized preprocessing and feature vector integration. The disease progression prediction model constructed using machine learning methods takes the patient's feature vector X as input and can efficiently predict the patient's disease progression fluctuation probability index Progress. Furthermore, combined with the correction coefficient algorithm and the set correction fluctuation threshold Cthe, the recovery period content extracted from the patient's initial medical record can be dynamically corrected to obtain the corrected recovery period Tad. Through feedback and iterative optimization, the calculation formula of the corrected survival period is continuously verified and updated, ultimately achieving the effect of improving prediction accuracy and adapting to individual differences in patients. It can effectively make up for the single prediction deficiency problem existing in the traditional method during the processing process and avoid prediction errors caused by ignoring individual differences in patients.

[0050] (2) By standardizing and preprocessing the parameters in the patient tumor identification data set P, the dimensional differences between different parameters can be eliminated. The construction of the feature vector X enables the individual features extracted from the patient data to be input into the disease progression prediction model in a consistent format. Combining the weighted linear combination z with the Sigmoid function, the patient's disease progression fluctuation probability index Progress can be accurately predicted. Through the training of the supervised learning algorithm and the optimization of the cross-entropy loss function, the prediction performance of the model has been continuously improved iteratively, which not only improves the prediction accuracy of the model, but also reduces the deviation caused by data inconsistency or errors, enhances the generalization ability of the model, and provides more reliable data support for clinical personalized treatment and decision-making.

[0051] (3) By comparing with the preset correction operation fluctuation threshold Cthe, the correction mechanism can be automatically triggered. When the correction coefficient Cm exceeds the threshold, the system will start the recovery period correction and then calculate the corrected recovery period Tad. This process makes the patient's treatment cycle prediction more personalized and accurate. The combination of the corrected recovery period Tad and the initial recovery period Tinit ensures the flexibility and adjustment space of the clinical treatment plan. In the further feedback and iterative optimization process, the system verifies and optimizes the corrected recovery period by comparing with the diagnostic recovery period data provided by the attending physician, ensuring the accuracy and reliability of the prediction model. Through this dynamic optimization, the accuracy of the prediction can be significantly improved, errors can be reduced, and the scientific nature of clinical decision-making can be improved, providing patients with more personalized and accurate treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic diagram of the steps of the data processing method for digestive tract tumors of the present invention;

[0053] Figure 2 This is a schematic block diagram of the data processing system for digestive tract tumors according to the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0055] Example 1

[0056] The present invention provides a data processing method for digestive tract tumors, please refer to Figure 1 , including the following steps:

[0057] S1. Collect and obtain patient tumor marker data from the patient's electronic health record source, imaging examination record source, blood test record source, and genomic test record source to form the patient's tumor marker data set P;

[0058] S2. Perform data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrate the processed patient tumor marker data set P to obtain a feature vector X;

[0059] S3. Build a disease progression prediction model using a machine learning method, use the feature vector X as the input feature of the disease progression prediction model, and obtain the disease progression fluctuation probability index Progress predicted for the patient through training;

[0060] S4. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm, and comparing it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result. Based on the correction trigger result, the recovery period content extracted from the patient's initial medical record is corrected to obtain a corrected recovery period Tad;

[0061] S5. Perform feedback and iterative optimization based on the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

[0062] In this embodiment, through steps S1 to S5, accurate prediction and dynamic adjustment of the patient's disease recovery progress are achieved. The patient's key physiological parameters and tumor characteristics are collected from multiple data sources to form the patient's tumor identification data set P. The consistency and computability of the data are ensured through standardized preprocessing and feature vector integration. The disease progression prediction model constructed using machine learning methods takes the patient's feature vector X as input and can efficiently predict the patient's disease progression fluctuation probability index Progress. Furthermore, combined with the correction coefficient algorithm and the set correction fluctuation threshold Cthe, the recovery period content extracted from the patient's initial medical record can be dynamically corrected to obtain the corrected recovery period Tad. Through feedback and iterative optimization, the calculation formula of the corrected survival period is continuously verified and updated, ultimately achieving the effect of improving prediction accuracy and adapting to individual differences among patients. It can effectively make up for the problem of insufficient single prediction in the processing process of traditional methods and avoid prediction errors caused by ignoring individual differences among patients.

[0063] Example 2

[0064] This embodiment is explained in Example 1, please refer to Figure 1 , specifically: the S1 includes S11 and S12;

[0065] S11. Collect individual difference-related parameters from clinical data sources, including extracting patient individual difference-related parameters from electronic health record sources, imaging examination record sources, blood test record sources, and genomic test record sources, including age A, sex S, weight W, height H, blood pressure BP1, gene mutation GM, and tumor size TS, to form a query collection data set CC = {A, S, W, H, BP, GM, TS};

[0066] Among them, age A, gender S, weight W and height H are obtained by entering the patient's unique information from the electronic health record source; tumor size TS is obtained by entering the patient's unique information from the imaging examination record source; blood pressure BP is obtained by entering the patient's unique information in the blood test record source to query the blood pressure measurement record; whether the gene is mutated GM is obtained by entering the patient's unique information in the genomic detection record source to obtain the gene sequence analysis results and mutation screening results.

[0067] S12, integrating the query-collected data set CC = {A, S, W, H, BP, GM, TS} to construct a patient tumor identification data set P;

[0068] Specifically, binary coding and categorical variable coding are performed on the sex S and whether the gene is mutated GM in the query data group CC to obtain the categorical variable sex Se and the binary gene mutation GMe;

[0069] Among them, the categorical variable sex Se includes classifying male as 1 and female as 2 to distinguish sex; the binary gene mutation GMe includes setting 0 as no gene mutation and 1 as the presence of gene mutation;

[0070] Substitute the categorical variable sex Se and the binary gene mutation GMe into the query data set CC = {A, S, W, H, BP, GM, TS} for replacement integration to obtain the integrated patient tumor identification data set P;

[0071] The patient tumor identification data set P = {A, Se, W, H, BP, GMe, TS}.

[0072] In this embodiment, by collecting individual difference-related parameters from multiple clinical data sources, it is possible to comprehensively acquire and integrate the patient's multidimensional health information, including age A, sex S, weight W, height H, blood pressure BP1, gene mutation status GM, and tumor size TS, to form a complete query data set C. By processing and encoding this data, particularly binary encoding of sex S and gene mutation GM, the raw data is converted into standardized categorical and binary variables, effectively improving the computational applicability and consistency of the data. In this process, the processing of specific parameters such as sex and gene mutation makes the data integration more consistent with the needs of subsequent analysis and model building. Through these processing, the patient's tumor marker data set P is formed, providing a more refined and adaptable data input foundation for subsequent personalized diagnosis and disease prediction. The advantage of this method is that it fully considers individual differences, not only improving data utilization efficiency but also laying a solid foundation for accurate model predictions, ensuring that personalized treatment plans can be adjusted according to the patient's specific characteristics, thereby avoiding bias caused by improper data processing and enhancing the accuracy and operability of predictions.

[0073] Example 3

[0074] This embodiment is explained in Example 2, please refer to Figure 1 Specifically: S2 includes S21;

[0075] S21. Perform data preprocessing on the patient's age A, weight W, height H, blood pressure BP, and tumor size TS in the tumor identification data set P, including standardization preprocessing, to obtain the preprocessed processed age Ac, processed weight Wc, processed height Hc, processed blood pressure BPc, and processed tumor size TSc, and substitute these data into the patient's tumor identification data set P for replacement, to obtain a feature vector X = {Ac, Se, Wc, Hc, BPc, GMe, TSc};

[0076] Among them, the standardization preprocessing eliminates the dimensional differences of the data by using feature scaling technology.

[0077] Said S3 includes S31 and S32;

[0078] S31. Construct a disease progression prediction model by using a machine learning method, and train the model using the feature vector X as an input feature to obtain a weighted linear combination z. Simultaneously, transform the weighted linear combination z using a Sigmoid function to obtain a predicted disease progression fluctuation probability index Progress for the patient.

[0079] The disease progression fluctuation probability index Progress is obtained by the following calculation formula:

[0080] z=(a1*Ac+a2*Se+a3*Wc+a4*Hc+a5*BPc+a6*GMe+a7*TSc+β);

[0081]

[0082] Where z represents a weighted linear combination, e represents a Sigmoid function, which is used to map the disease progression fluctuation probability index Progress to the range of 0 to 1, a1, a2, a3, a4, a5, a6, and a7 represent the weight coefficients of treatment age Ac, categorical variable sex Se, treatment weight Wc, treatment height Hc, treatment blood pressure BPc, binary gene mutation GMe, and treatment tumor size TSc, respectively, β represents a bias term, and a1+a2+a3+a4+a5+a6+a7=1. The specific value is set by the user.

[0083] S32. Use a supervised learning algorithm to find the optimal weight coefficient and bias term β for the disease progression prediction model, including using a cross-entropy loss function to measure the difference between the prediction and the actual value, obtaining the total loss L(ai, β) of the cross-entropy loss function, and then using a gradient descent algorithm to gradually adjust the weight coefficient and bias term β to minimize the total loss L(ai, β) to iteratively optimize the constructed disease progression prediction model. The smaller the total loss L(ai, β), the better the model prediction result.

[0084] The total loss L(ai, β) is obtained by the following calculation formula:

[0085]

[0086] Where m represents the number of samples in the training set, specifically how many patient data are used to train the disease progression prediction model, yj represents the true label of the j-th patient, and the value of yj is 0 or 1. When the j-th patient has disease progression, yj = 1, otherwise yj = 0. It is specifically obtained by extracting the training data set. P(Progressj) represents the disease progression fluctuation probability index of the j-th patient output by the disease progression prediction model, ranging from 0 to 1. Log represents the logarithmic function, which is used to convert the probability value into the amount of information in the loss function. By taking the logarithm of the predicted probability, an incorrect prediction will result in a larger loss. For example, the predicted probability is close to 0 and the true label is 1.

[0087] In this embodiment, by standardizing and preprocessing the key parameters in the patient tumor identification data set P, such as the processed age Ac, weight Wc, height Hc, blood pressure BPc, and tumor size TSc, the dimensional differences between different parameters can be eliminated, providing a unified scale data input for subsequent model training. The construction of the feature vector X enables the individual features extracted from the patient data to be input into the disease progression prediction model in a consistent format, enhancing the model's adaptability and computing power for various types of data. When using a machine learning method to construct a disease progression prediction model, the feature vector X is used as input, combined with the weighted linear combination z and the Sigmoid function, to accurately predict the patient's disease progression fluctuation probability index Progress, and map the model's output to a range of 0 to 1, ensuring the interpretability of the prediction results and the practicality of clinical application. Through the training of the supervised learning algorithm and the optimization of the cross-entropy loss function, the model's prediction performance has been continuously iteratively improved, so that the weight coefficient and bias term β are optimized and adjusted, and the disease progression prediction model finally constructed can achieve more accurate disease progression prediction. Compared with traditional methods, this process not only improves the prediction accuracy of the model, but also reduces the deviation caused by data inconsistency or errors, enhances the generalization ability of the model, and provides more reliable data support for clinical personalized treatment and decision-making.

[0088] Example 4

[0089] This embodiment is explained in Example 3, please refer to Figure 1 Specifically: S4 includes S41 and S42;

[0090] S41. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm;

[0091] The correction coefficient Cm is obtained by the following calculation formula:

[0092]

[0093] Wherein, B represents the exponential power, which is specifically used to adjust the nonlinear effect of disease progression on the correction coefficient Cm, γ represents the effect coefficient of age A on the correction coefficient Cm, c1, c2 and c3 represent the effect coefficients on the correction coefficient Cm, specifically c1 represents the effect coefficient of disease progression fluctuation probability index Progress on the correction coefficient Cm, c2 represents the effect coefficient of age A on the correction coefficient Cm, c3 represents the effect coefficient of weight W on the correction coefficient Cm, c4 represents the nonlinear effect coefficient of weight W on the correction coefficient Cm, exp represents the exponential function, SZ represents the body mass index, specifically through Calculation formula is obtained, W represents weight, H represents height.

[0094] S42. Compare the correction coefficient Cm with a preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and correct the recovery period extracted from the patient's initial medical record based on the correction trigger result to obtain an initial recovery period Tinit. Combined with the correction coefficient Cm, a corrected recovery period Tad is obtained after correction. The initial recovery period Tinit is obtained from the treatment cycle extracted from the initial medical record evaluated by the patient's attending physician.

[0095] The correction trigger result is obtained by the following comparison method:

[0096] When the correction coefficient Cm is less than the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as a non-trigger result;

[0097] When the correction coefficient Cm ≥ the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as the trigger result;

[0098] The modified recovery period Tad is obtained by the calculation formula Tad=Tinit*(1+Cm).

[0099] The S5 includes S51;

[0100] S51. Feedback and iterative optimization are performed based on the revised recovery period Tad, including verification and updating of the revised recovery period Tad. Verification includes comparing the revised recovery period Tad with the diagnostic recovery period data provided by the patient's attending physician to check the accuracy of the revised recovery period Tad and recording the recovery period difference ΔTad. When the recovery period difference ΔTad is less than 5%, the revised recovery period Tad is not optimized or updated. When the recovery period difference ΔTad is greater than or equal to 5%, the revised recovery period Tad is optimized and updated.

[0101] Among them, the optimization update includes using the recovery period difference △Tad to proportionally adjust the influence coefficient of the correction coefficient Cm and the weight coefficient of the disease progression prediction model.

[0102] In this embodiment, by introducing a correction coefficient Cm and combining it with the disease progression fluctuation probability index Progress, patient age A, and weight W, the patient's recovery period prediction model can be precisely adjusted. The calculation of the correction coefficient Cm not only takes into account the nonlinear effects of disease progression but also incorporates individual differences such as age, weight, and body mass index. This allows for more flexible adaptation to the physiological characteristics of different patients and optimizes recovery period prediction results. By comparing the correction coefficient Cm with a preset correction operation fluctuation threshold Cthe, a correction mechanism is automatically triggered. When the correction coefficient Cm exceeds the threshold, the system initiates recovery period correction and calculates the corrected recovery period Tad. This process makes the patient's treatment cycle prediction more personalized and accurate. The combination of the corrected recovery period Tad with the initial recovery period Tinit ensures flexibility and adjustment in clinical treatment plans. During further feedback and iterative optimization, the system verifies and optimizes the corrected recovery period by comparing it with the diagnostic recovery period data provided by the attending physician, ensuring the accuracy and reliability of the prediction model. This dynamic optimization can significantly improve prediction accuracy, reduce errors, and enhance the scientific nature of clinical decision-making, providing patients with more personalized and precise treatment plans.

[0103] Example 5

[0104] For data processing systems for digestive tract tumors, please refer to Figure 2 ,Specifically: including data acquisition module, data preprocessing module, modeling and analysis module, correction and evaluation module and iterative optimization module;

[0105] The data acquisition module collects data from the patient's electronic health record source, imaging examination record source, blood test record source and genomic test record source to form the patient's tumor identification data set P;

[0106] The data preprocessing module performs data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrates the processed patient tumor marker data set P to obtain a feature vector X;

[0107] The modeling and analysis module constructs a disease progression prediction model by using a machine learning method, and uses the feature vector X as the input feature of the disease progression prediction model, and obtains the disease progression fluctuation probability index Progress predicted for the patient through training;

[0108] The correction evaluation module substitutes the acquired disease progression fluctuation probability index Progress into the correction coefficient algorithm formula to perform calculations to obtain the correction coefficient Cm, and compares it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and corrects the recovery period content extracted from the patient's initial medical record according to the correction trigger result to obtain a corrected recovery period Tad.

[0109] The iterative optimization module performs feedback and iterative optimization according to the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

[0110] While embodiments of the present invention have been shown and described, it will be appreciated 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 defined by the appended claims and their equivalents.

Claims

1. A data processing method for digestive tract tumors, characterized by: The following steps are involved: S1. Collect and obtain patient tumor marker data from the patient's electronic health record source, imaging examination record source, blood test record source, and genomic test record source to form the patient's tumor marker data set P; S2. Perform data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrate the processed patient tumor marker data set P to obtain a feature vector X; S3. Build a disease progression prediction model using a machine learning method, use the feature vector X as the input feature of the disease progression prediction model, and obtain the disease progression fluctuation probability index Progress predicted for the patient through training; Said S3 includes S31 and S32; S31. Construct a disease progression prediction model by using a machine learning method, and train the model using the feature vector X as an input feature to obtain a weighted linear combination z. Simultaneously, transform the weighted linear combination z using a Sigmoid function to obtain a predicted disease progression fluctuation probability index Progress for the patient. The disease progression fluctuation probability index Progress is obtained by the following calculation formula: z=(a1*Ac+a2*Se+a3*Wc+a4*Hc+a5*BPc+a6*GMe+a7*TSc+β); Where z represents a weighted linear combination, e represents a Sigmoid function, which is used to map the disease progression fluctuation probability index Progress to the range of 0 to 1, a1, a2, a3, a4, a5, a6, and a7 represent the weight coefficients of treatment age Ac, categorical variable sex Se, treatment weight Wc, treatment height Hc, treatment blood pressure BPc, binary gene mutation GMe, and treatment tumor size TSc, respectively, β represents a bias term, and a1+a2+a3+a4+a5+a6+a7=1. The specific value is set by the user. S32. Use a supervised learning algorithm to find the optimal weight coefficient and bias term β for the disease progression prediction model, including using a cross-entropy loss function to measure the difference between the predicted and actual values, obtaining the total loss of the cross-entropy loss function, and then using a gradient descent algorithm to gradually adjust the weight coefficient and bias term β to minimize the total loss, thereby iteratively optimizing the constructed disease progression prediction model; S4. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm, and comparing it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result. Based on the correction trigger result, the recovery period content extracted from the patient's initial medical record is corrected to obtain a corrected recovery period Tad; Said S4 includes S41 and S42; S41. Substituting the obtained disease progression fluctuation probability index Progress into the correction coefficient algorithm formula for calculation to obtain the correction coefficient Cm; The correction coefficient Cm is obtained by the following calculation formula: Where B represents the exponential power, γ represents the influence coefficient of age A on the correction coefficient Cm, c1, c2 and c3 represent the influence coefficients on the correction coefficient Cm, c4 represents the nonlinear influence coefficient of weight W on the correction coefficient Cm, exp represents the exponential function, SZ represents the body mass index, and is specifically expressed by Calculation formula is obtained, W represents weight, H represents height; S42. Compare the correction coefficient Cm with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and correct the recovery period extracted from the patient's initial medical record according to the correction trigger result to obtain the initial recovery period Tinit. Combined with the correction coefficient Cm, a corrected recovery period Tad is obtained. The correction trigger result is obtained by the following comparison method: When the correction coefficient Cm is less than the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as a non-trigger result; When the correction coefficient Cm ≥ the correction operation fluctuation threshold Cthe, the correction trigger result is obtained as the trigger result; The modified recovery period Tad is obtained by the calculation formula Tad=Tinit*(1+Cm); S5. Perform feedback and iterative optimization based on the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

2. The data processing method for digestive tract tumors according to claim 1, characterized in that: Said S1 includes S11 and S12; S11. Collect individual difference-related parameters from clinical data sources, including extracting patient individual difference-related parameters from electronic health record sources, imaging examination record sources, blood test record sources, and genomic test record sources, including age A, sex S, weight W, height H, blood pressure BP1, gene mutation GM, and tumor size TS, to form a query collection data set CC = {A, S, W, H, BP, GM, TS}; Among them, age A, gender S, weight W and height H are obtained by entering the patient's unique information from the electronic health record source; tumor size TS is obtained by entering the patient's unique information from the imaging examination record source; blood pressure BP is obtained by entering the patient's unique information in the blood test record source to query the blood pressure measurement record; whether the gene is mutated GM is obtained by entering the patient's unique information in the genomic detection record source to obtain the gene sequence analysis results and mutation screening results.

3. The data processing method for digestive tract tumors according to claim 2, characterized in that: S12, integrating the query-collected data set CC = {A, S, W, H, BP, GM, TS} to construct a patient tumor identification data set P; Specifically, binary coding and categorical variable coding are performed on the sex S and whether the gene is mutated GM in the query data group CC to obtain the categorical variable sex Se and the binary gene mutation GMe; Among them, the categorical variable sex Se includes classifying male as 1 and female as 2 to distinguish sex; the binary gene mutation GMe includes setting 0 as no gene mutation and 1 as the presence of gene mutation; Substitute the categorical variable sex Se and the binary gene mutation GMe into the query data set CC = {A, S, W, H, BP, GM, TS} for replacement integration to obtain the integrated patient tumor identification data set P; The patient tumor identification data set P = {A, Se, W, H, BP, GMe, TS}.

4. The data processing method for digestive tract tumors according to claim 3, characterized in that: Said S2 includes S21; S21. Perform data preprocessing on the patient's age A, weight W, height H, blood pressure BP, and tumor size TS in the tumor identification data set P, including standardization preprocessing, to obtain the preprocessed processed age Ac, processed weight Wc, processed height Hc, processed blood pressure BPc, and processed tumor size TSc, and substitute these data into the patient's tumor identification data set P for replacement, to obtain a feature vector X = {Ac, Se, Wc, Hc, BPc, GMe, TSc}; Among them, the standardization preprocessing eliminates the dimensional differences of the data by using feature scaling technology.

5. The data processing method for digestive tract tumors according to claim 1, characterized in that: Said S5 includes S51; S51. Feedback and iterative optimization are performed based on the revised recovery period Tad, including verification and updating of the revised recovery period Tad. Verification includes comparing the revised recovery period Tad with the diagnostic recovery period data provided by the patient's attending physician to check the accuracy of the revised recovery period Tad and recording the recovery period difference ΔTad. When the recovery period difference ΔTad is less than 5%, the revised recovery period Tad is not optimized or updated. When the recovery period difference ΔTad is greater than or equal to 5%, the revised recovery period Tad is optimized and updated. The optimization update includes proportionally adjusting the influence coefficient of the correction coefficient Cm and the weight coefficient of the disease progression prediction model using the recovery period difference △Tad.

6. A data processing system for digestive tract tumors, applied to the data processing method for digestive tract tumors according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, data preprocessing module, modeling and analysis module, correction and evaluation module and iterative optimization module; The data acquisition module collects data from the patient's electronic health record source, imaging examination record source, blood test record source and genomic test record source to form the patient's tumor identification data set P; The data preprocessing module performs data preprocessing on the patient tumor marker data set P, including standardization preprocessing, and then integrates the processed patient tumor marker data set P to obtain a feature vector X; The modeling and analysis module constructs a disease progression prediction model by using a machine learning method, and uses the feature vector X as the input feature of the disease progression prediction model, and obtains the disease progression fluctuation probability index Progress predicted for the patient through training; The correction evaluation module substitutes the acquired disease progression fluctuation probability index Progress into the correction coefficient algorithm formula to perform calculations to obtain the correction coefficient Cm, and compares it with the preset correction operation fluctuation threshold Cthe to obtain a correction trigger result, and corrects the recovery period content extracted from the patient's initial medical record according to the correction trigger result to obtain a corrected recovery period Tad. The iterative optimization module performs feedback and iterative optimization according to the modified recovery period Tad, including verifying and updating the modified recovery period Tad.

Images