Prostate cancer noninvasive detection method and system based on clinical data of patient

Abstract

The invention discloses a prostate cancer noninvasive detection method and system based on patient clinical data, and relates to the technical field of medical detection, and the method comprises the following operation steps: S1, data acquisition and preparation; s2, constructing a model; s3, model verification; s4, performing model evaluation; and S5: feature mapping and result interpretation. According to the prostate cancer non-invasive detection method and system based on the clinical data of the patient, a non-invasive method is adopted, such as urine detection or detection of biomarkers in blood, harm to the body of the patient is reduced, comprehensive clinical data including PSA level, urine detection results, imaging examination data and the like are utilized, and the prostate cancer non-invasive detection method and system based on the clinical data of the patient are obtained. A machine learning algorithm is combined for comprehensive analysis, more accurate risk assessment is provided, meanwhile, various machine learning models such as logistic regression, a support vector machine and deep learning are constructed, and the models are used for training and predicting clinical data, so that the risk of prostatic cancer is assessed.

Description

Technical Field

[0001] The present invention relates to the field of medical detection technology, and in particular to a non-invasive detection method and system for prostate cancer based on patient clinical data. Background Art

[0002] Prostate cancer is one of the most common malignant tumors in men. Early diagnosis is crucial to improving patient prognosis. Traditional prostate cancer detection methods mainly include prostate-specific antigen (PSA) testing, digital rectal examination (DRE) and prostate puncture biopsy. However, these methods have certain limitations, which has promoted the research on non-invasive detection methods for prostate cancer.

[0003] Among them, the prostate-specific antigen (PSA) test: The PSA test assesses the risk of prostate cancer by measuring the PSA level in the blood. PSA is a protein secreted by the prostate. Its increased level may be related to prostate cancer. Its disadvantages are: increased PSA levels may also be caused by benign prostatic hyperplasia (BPH) or other non-cancerous diseases, so there are false positives and false negatives, causing unnecessary anxiety for patients or missing real cancer cases; Digital rectal examination (DRE): DRE is a procedure in which doctors examine the size, shape, and texture of the prostate through rectal palpation to detect possible abnormalities. However, its disadvantages are: DRE has limited sensitivity and specificity, making it difficult to accurately detect early prostate cancer. Among them, prostate puncture biopsy: through transrectal ultrasound-guided puncture sampling, prostate tissue is obtained for pathological examination to confirm the presence of prostate cancer. Its disadvantages are: this method is invasive and may cause patient discomfort and complications. At the same time, the biopsy may not cover all cancerous areas and may miss some cancer cases. Although existing prostate cancer detection methods are widely used in clinical applications, they still have the following shortcomings: Invasiveness: Prostate puncture biopsy is invasive and may cause complications such as infection and bleeding, which imposes a certain burden on the patient's body; False positive and false negative rates: PSA testing and DRE have problems with false positives and false negatives, which may lead to unnecessary follow-up examinations or miss true cancer cases, affecting the accuracy of diagnosis; Lack of comprehensive consideration: Existing detection methods are usually performed in isolation and fail to fully combine the patient's comprehensive clinical data (such as family history, genetic information, metabolite levels, etc.) for comprehensive evaluation; Insufficient data integration and analysis: Existing technologies lack advanced algorithms and models in data analysis and comprehensive risk assessment, and fail to make full use of big data and machine learning technologies to improve prediction accuracy.

[0004] Therefore, it is necessary to propose a non-invasive detection method and system for prostate cancer based on patient clinical data to solve the above problems. Summary of the invention

[0005] The main purpose of the present invention is to provide a non-invasive detection method and system for prostate cancer based on patient clinical data, which can effectively solve the problems in the background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is: A non-invasive detection method for prostate cancer based on patient clinical data, comprising the following steps: S1: Data collection and preparation: This includes data collection and preparation, where data collection is used to collect clinical data of patients, including but not limited to prostate-specific antigen levels, urine test results, imaging examination data, family history, age, and lifestyle of patients, where preparation includes preprocessing of collected clinical data of patients, and preprocessing includes cleaning and preprocessing of collected data, including missing value filling, data standardization, and outlier processing; S2: Model construction: including model selection and model training, where model selection: selecting a variety of machine learning models for training, including logistic regression, support vector machine, random forest, gradient boosting machine and deep learning model; Model training: Select a suitable machine learning model and use the training data to train the model. Adjust and optimize the model parameters to ensure that it performs well on the training data. Use cross-validation techniques to evaluate the generalization ability of the model during the training process. S3: Model validation: External validation: Validation is performed on an independent validation dataset that is different from the training set to evaluate the generalization ability and stability of the model. The reliability of the model is confirmed by comparing the prediction results of the external dataset with the actual situation. Cross-validation: Use K-fold cross-validation technology when dividing the data set, divide the data set into K subsets, use K-1 subsets for training each time, and use the remaining subset for testing, so as to evaluate the performance of the model multiple times and reduce the performance fluctuation caused by the accidental data division; Overfitting check: Check whether the model is overfitting. You can avoid overfitting by checking the difference in loss function between the training set and the validation set, using regularization techniques, and choosing an appropriate number of training times. Model robustness verification: Verify the robustness of the model by testing noise data, extreme value data, and different subgroups, including different age groups and different disease processes, to ensure its wide applicability; S4: Model evaluation: including drawing ROC curves and evaluation indicators, where the ROC curve is used to select the best model and determine the optimal decision threshold; Evaluation indicators: The performance of each model is evaluated by drawing the ROC curve, i.e. the receiver operating characteristic curve, and the AUC value of each model, i.e. the area under the curve, is calculated to determine the accuracy, sensitivity and specificity of the model at different thresholds. An AUC value of the ROC curve of 0.8 or above indicates that the model has good classification performance; S5: Feature mapping and result interpretation: Feature mapping: The feature mapping algorithm is used to combine the prediction results of the machine learning model with the clinical characteristics of the patient to form a comprehensive risk assessment tool. The feature mapping algorithm is used to convert the model's predicted probability into an easily interpretable clinical risk score; Interpretation of results: Provide personalized advice and follow-up examination recommendations based on risk scores, combine risk scores with actual clinical conditions, and help doctors and patients make decisions.

[0007] Preferably, in step S1, the imaging examination data is an ultrasonic examination result.

[0008] A non-invasive detection system for prostate cancer based on patient clinical data is applied to the above-mentioned non-invasive detection method for prostate cancer based on patient clinical data, including a main control unit, the main control unit includes a data acquisition unit, the output end of the data acquisition unit is electrically connected to the input end of a training module, the output end of the training module is electrically connected to the input end of a model evaluation unit, and the input end of the model evaluation unit is electrically connected to the output end of a feature mapping unit.

[0009] Preferably, the data acquisition unit comprises a patient data acquisition module, and the output end of the patient data acquisition module is electrically connected to the input end of the data preprocessing module; The patient data acquisition module is used to acquire the patient's clinical data, and the data preprocessing module is used to preprocess the acquired clinical data, including data cleaning, missing value filling, data standardization and outlier processing.

[0010] Preferably, the input end of the training module is electrically connected to the output end of the model selection module, the input end of the model selection module is electrically connected to the input end of the model construction unit, the model evaluation unit includes an ROV curve drawing module, the output end of the ROV curve drawing module is electrically connected to the input end of the AUC value calculation module, and the output end of the AUC value calculation module is electrically connected to the input end of the judgment module; The model construction unit is used to construct a machine learning model, the model selection module is used to select a required learning module, the ROV curve drawing module is used to construct a ROC curve, the AUC value calculation module is used to calculate the AUC value, and the judgment module judges the accuracy, sensitivity and specificity of the model at different thresholds based on the constructed ROC curve and AUC value.

[0011] Preferably, the feature mapping unit includes a data quenching module, the output end of the data quenching module is electrically connected to the input end of the risk assessment module, the output end of the risk assessment module is electrically connected to the input end of the scoring module, the output end of the scoring module is electrically connected to the input end of the diagnosis and treatment suggestion generation module, and the input end of the diagnosis and treatment suggestion generation module is electrically connected to the output end of the hospital database; The data quenching module combines the prediction results of the machine learning model with the clinical characteristics of the patient based on a feature mapping algorithm. The risk assessment module assesses the risk based on the combined data. The scoring module performs scoring based on the content of the assessment. The diagnosis and treatment recommendation generation module generates diagnosis and treatment recommendations based on the content of the assessment and scoring. Beneficial Effects

[0012] Compared with the prior art, the present invention provides a non-invasive detection method and system for prostate cancer based on patient clinical data, which has the following beneficial effects: This non-invasive prostate cancer detection method and system based on patient clinical data adopts non-invasive methods, such as urine testing or biomarker testing in the blood, to reduce physical harm to patients. It uses comprehensive clinical data, including PSA levels, urine test results, imaging examination data, etc., combined with machine learning algorithms for comprehensive analysis to provide more accurate risk assessments. At the same time, it constructs a variety of machine learning models, such as logistic regression, support vector machines, deep learning, etc., and uses these models to train and predict clinical data to assess the risk of prostate cancer.

[0013] This non-invasive detection method and system for prostate cancer based on patient clinical data evaluates the prediction accuracy of each machine learning model by drawing the ROC curve, that is, the receiver operating characteristic curve. The ROC curve can effectively display the sensitivity and specificity of the model at different thresholds, and help determine the optimal performance and discrimination ability of the model. The ROC curve can intuitively display the performance of the model at different thresholds, so as to select the best model. By drawing and analyzing the ROC curve, the model selection and threshold setting are optimized, and the reliability of risk prediction is improved.

[0014] This non-invasive detection method and system for prostate cancer based on patient clinical data. In the prior art, although there are marker detection methods based on urine or blood, the sensitivity and specificity of these methods are often insufficient, and most methods fail to effectively combine clinical data for comprehensive evaluation. Through the above improvements, the present invention aims to provide a more accurate, non-invasive, and comprehensive prostate cancer detection method to overcome the shortcomings of the prior art and improve the effectiveness of early diagnosis and treatment of prostate cancer.

[0015] This non-invasive prostate cancer detection method and system based on patient clinical data avoids the invasiveness of traditional puncture biopsy, reduces physical harm to patients and the risk of complications, collects data in a non-invasive manner, reduces the pain and complexity of testing, and improves patient acceptance.

[0016] This non-invasive detection method and system for prostate cancer based on patient clinical data has built a model pool containing multiple machine learning models, and uses different algorithms to train and predict clinical data. This method can capture complex patterns in the data better than a single model, and improves the accuracy of risk prediction. By combining multiple models such as logistic regression, support vector machine (SVM), random forest, deep learning, etc., it comprehensively improves the prediction performance and overcomes the limitations that may exist in a single model. At the same time, it comprehensively analyzes multiple data sources to provide more accurate risk assessment and help doctors better understand the patient's condition.

[0017] This non-invasive detection method and system for prostate cancer based on patient clinical data combines the prediction results of the machine learning model with clinical characteristics through a proposed feature mapping algorithm, thereby forming a more intuitive risk assessment tool. Through feature mapping, the prediction results and clinical data are effectively integrated, providing a risk assessment method that is easy to interpret and apply. The present invention not only improves the accuracy of non-invasive detection of prostate cancer, but also optimizes the data analysis process, providing a more advanced and comprehensive detection method for clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the present invention; Figure 2 It is a data-driven AI disease perception model diagram of the present invention; Figure 3 It is a system block diagram of the main control unit of the present invention. DETAILED DESCRIPTION

[0019] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0020] like Figure 1-Figure 3 As shown, a non-invasive detection method for prostate cancer based on patient clinical data includes the following steps: S1: Data collection and preparation: This includes data collection and preparation, where data collection is used to collect clinical data of patients, including but not limited to prostate-specific antigen levels, urine test results, imaging examination data, family history, age, and lifestyle of patients, where preparation includes preprocessing of collected clinical data of patients, and preprocessing includes cleaning and preprocessing of collected data, including missing value filling, data standardization, and outlier processing; S2: Model construction: including model selection and model training, where model selection: selecting a variety of machine learning models for training, including logistic regression, support vector machine, random forest, gradient boosting machine and deep learning model; Model training: Select a suitable machine learning model and use the training data to train the model. Adjust and optimize the model parameters to ensure that it performs well on the training data. Use cross-validation techniques to evaluate the generalization ability of the model during the training process. S3: Model Validation: External Validation: Validation is performed on an independent validation dataset that is different from the training set to evaluate the generalization ability and stability of the model. The reliability of the model is confirmed by comparing the prediction results of the external dataset with the actual situation.

[0021] S3: Model validation: External validation: Validation is performed on an independent validation dataset that is different from the training set to evaluate the generalization ability and stability of the model. The reliability of the model is confirmed by comparing the prediction results of the external dataset with the actual situation. Cross-validation: Use K-fold cross-validation technology when dividing the data set, divide the data set into K subsets, use K-1 subsets for training each time, and use the remaining subset for testing, so as to evaluate the performance of the model multiple times and reduce the performance fluctuation caused by the accidental data division; Overfitting check: Check whether the model is overfitting. You can avoid overfitting by checking the difference in loss function between the training set and the validation set, using regularization techniques, and choosing an appropriate number of training times. Model robustness verification: The robustness of the model is verified by testing noisy data, extreme value data, and different sub-groups, including different age groups and different disease processes, to ensure its wide applicability.

[0022] S4: Model evaluation: including drawing ROC curves and evaluation indicators, where the ROC curve is used to select the best model and determine the optimal decision threshold; Evaluation indicators: The performance of each model is evaluated by drawing the ROC curve, i.e. the receiver operating characteristic curve, and the AUC value of each model, i.e. the area under the curve, is calculated to determine the accuracy, sensitivity and specificity of the model at different thresholds. An AUC value of the ROC curve of 0.8 or above indicates that the model has good classification performance; S5: Feature mapping and result interpretation: Feature mapping: The feature mapping algorithm is used to combine the prediction results of the machine learning model with the clinical characteristics of the patient to form a comprehensive risk assessment tool. The feature mapping algorithm is used to convert the model's predicted probability into an easily interpretable clinical risk score; Interpretation of results: Provide personalized advice and follow-up examination recommendations based on risk scores, combine risk scores with actual clinical conditions, and help doctors and patients make decisions.

[0023] like Figure 1-Figure 3 As shown, a non-invasive detection system for prostate cancer based on patient clinical data is applied to the above-mentioned non-invasive detection method for prostate cancer based on patient clinical data, including a main control unit, the main control unit includes a data acquisition unit, the output end of the data acquisition unit is electrically connected to the input end of the training module, the output end of the training module is electrically connected to the input end of the model evaluation unit, and the input end of the model evaluation unit is electrically connected to the output end of the feature mapping unit.

[0024] The data acquisition unit includes a patient data acquisition module, the output end of the patient data acquisition module is electrically connected to the input end of the data preprocessing module, the patient data acquisition module is used to collect the patient's clinical data, and the data preprocessing module is used to preprocess the collected clinical data, including data cleaning, missing value filling, data standardization and outlier processing.

[0025] The input end of the training module is electrically connected to the output end of the model selection module, the input end of the model selection module is electrically connected to the input end of the model construction unit, the model evaluation unit includes an ROV curve drawing module, the output end of the ROV curve drawing module is electrically connected to the input end of the AUC value calculation module, the output end of the AUC value calculation module is electrically connected to the input end of the judgment module, the model construction unit is used to construct a machine learning model, the model selection module is used to select the required learning module, the ROV curve drawing module is used to construct a ROC curve, the AUC value calculation module is used to calculate the AUC value, and the judgment module judges the accuracy, sensitivity and specificity of the model at different thresholds based on the constructed ROC curve and AUC value.

[0026] The feature mapping unit includes a data quenching module, the output end of the data quenching module is electrically connected to the input end of the risk assessment module, the output end of the risk assessment module is electrically connected to the input end of the scoring module, the output end of the scoring module is electrically connected to the input end of the diagnosis and treatment suggestion generation module, the input end of the diagnosis and treatment suggestion generation module is electrically connected to the output end of the hospital database, the data quenching module combines the prediction results of the machine learning model with the clinical characteristics of the patient based on the feature mapping algorithm, the risk assessment module performs risk assessment based on the combined data, the scoring module performs scoring processing based on the content of the assessment, and the diagnosis and treatment suggestion generation module generates diagnosis and treatment suggestions based on the content of the assessment and scoring.

[0027] It should be noted that the present invention is a non-invasive detection method and system for prostate cancer based on patient clinical data.

[0028] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A non-invasive detection method for prostate cancer based on patient clinical data, characterized in that: The steps include: S1: Data collection and preparation: This includes data collection and preparation, where data collection is used to collect clinical data of patients, including but not limited to prostate-specific antigen levels, urine test results, imaging examination data, family history, age, and lifestyle of patients, where preparation includes preprocessing of collected clinical data of patients, and preprocessing includes cleaning and preprocessing of collected data, including missing value filling, data standardization, and outlier processing; S2: Model construction: including model selection and model training, where model selection: selecting a variety of machine learning models for training, including logistic regression, support vector machine, random forest, gradient boosting machine and deep learning model; Model training: Select a suitable machine learning model and use the training data to train the model. Adjust and optimize the model parameters to ensure that it performs well on the training data. Use cross-validation techniques to evaluate the generalization ability of the model during the training process. S3: Model validation: External validation: Validation is performed on an independent validation dataset that is different from the training set to evaluate the generalization ability and stability of the model. The reliability of the model is confirmed by comparing the prediction results of the external dataset with the actual situation. Cross-validation: Use K-fold cross-validation technology when dividing the data set, divide the data set into K subsets, use K-1 subsets for training each time, and use the remaining subset for testing, so as to evaluate the performance of the model multiple times and reduce the performance fluctuation caused by the accidental data division; Overfitting check: Check whether the model is overfitting. You can avoid overfitting by checking the difference in loss function between the training set and the validation set, using regularization techniques, and choosing an appropriate number of training times. Model robustness verification: Verify the robustness of the model by testing noise data, extreme value data, and different subgroups, including different age groups and different disease processes, to ensure its wide applicability; S4: Model evaluation: including drawing ROC curves and evaluation indicators, where the ROC curve is used to select the best model and determine the optimal decision threshold; Evaluation indicators: The performance of each model is evaluated by drawing the ROC curve, i.e. the receiver operating characteristic curve, and the AUC value of each model, i.e. the area under the curve, is calculated to determine the accuracy, sensitivity and specificity of the model at different thresholds. An AUC value of the ROC curve of 0.8 or above indicates that the model has good classification performance; S5: Feature mapping and result interpretation: Feature mapping: The feature mapping algorithm is used to combine the prediction results of the machine learning model with the clinical characteristics of the patient to form a comprehensive risk assessment tool. The feature mapping algorithm is used to convert the model's predicted probability into an easily interpretable clinical risk score; Interpretation of results: Provide personalized advice and follow-up examination recommendations based on risk scores, combine risk scores with actual clinical conditions, and help doctors and patients make decisions.

2. The non-invasive detection method for prostate cancer based on patient clinical data according to claim 1, characterized in that: In step S1, the imaging examination data is the ultrasonic examination result.

3. A non-invasive detection system for prostate cancer based on patient clinical data, applied to the above-mentioned non-invasive detection method for prostate cancer based on patient clinical data, comprising a main control unit, characterized in that: The main control unit includes a data acquisition unit, the output end of the data acquisition unit is electrically connected to the input end of the training module, the output end of the training module is electrically connected to the input end of the model evaluation unit, and the input end of the model evaluation unit is electrically connected to the output end of the feature mapping unit.

4. The non-invasive prostate cancer detection system based on patient clinical data according to claim 3, characterized in that: The data acquisition unit comprises a patient data acquisition module, the output end of the patient data acquisition module is electrically connected to the input end of the data preprocessing module; The patient data acquisition module is used to acquire the patient's clinical data, and the data preprocessing module is used to preprocess the acquired clinical data, including data cleaning, missing value filling, data standardization and outlier processing.

5. The non-invasive prostate cancer detection system based on patient clinical data according to claim 3, characterized in that: The input end of the training module is electrically connected to the output end of the model selection module, the input end of the model selection module is electrically connected to the input end of the model construction unit, the model evaluation unit includes an ROV curve drawing module, the output end of the ROV curve drawing module is electrically connected to the input end of the AUC value calculation module, and the output end of the AUC value calculation module is electrically connected to the input end of the judgment module; The model construction unit is used to construct a machine learning model, the model selection module is used to select a required learning module, the ROV curve drawing module is used to construct a ROC curve, the AUC value calculation module is used to calculate the AUC value, and the judgment module judges the accuracy, sensitivity and specificity of the model at different thresholds based on the constructed ROC curve and AUC value.

6. The non-invasive prostate cancer detection system based on patient clinical data according to claim 3, characterized in that: The feature mapping unit includes a data quenching module, an output end of the data quenching module is electrically connected to an input end of a risk assessment module, an output end of the risk assessment module is electrically connected to an input end of a scoring module, an output end of the scoring module is electrically connected to an input end of a diagnosis and treatment suggestion generation module, and an input end of the diagnosis and treatment suggestion generation module is electrically connected to an output end of a hospital database; The data quenching module combines the prediction results of the machine learning model with the clinical characteristics of the patient based on a feature mapping algorithm. The risk assessment module assesses the risk based on the combined data. The scoring module performs scoring based on the content of the assessment. The diagnosis and treatment recommendation generation module generates diagnosis and treatment recommendations based on the content of the assessment and scoring.

Images