Tumor recurrence risk assessment and early warning system based on data analysis

Abstract

The invention discloses a tumor recurrence risk assessment and early warning system based on data analysis, and belongs to the technical field of medical information. The invention discloses a tumor recurrence risk assessment and early warning system based on data analysis. The system comprises a multi-data collection module, a big data processing platform, a model architecture system and an individual data processing system. According to the invention, the problems of lack of big data support, limited individualized prediction ability and insufficient real-time performance and continuous monitoring in the prior art are solved; according to the method, large-scale historical case data are integrated, the rich data sets are used for model training, the accuracy and generalization ability of a prediction model can be remarkably improved, multiple data sources such as genomics information, living habits and environmental factors are integrated, personalized risk assessment is provided for each patient, and the risk assessment efficiency is improved. The real-time monitoring interface is connected with various intelligent medical devices to collect the daily physiological parameter change condition of the patient, and continuous health management suggestions are provided for the tumor postoperative patient.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a tumor recurrence risk assessment and early warning system based on data analysis. Background Art

[0002] Tumor recurrence is an important problem after cancer treatment, which seriously affects the survival rate and quality of life of patients. Traditional tumor recurrence risk assessment methods mainly rely on the experience of clinicians and regular follow-up examinations, but this method is inefficient and difficult to achieve personalized and accurate risk prediction. With the development of big data and machine learning technology, tumor recurrence risk assessment and early warning systems based on data analysis have emerged. They aim to provide more accurate risk assessments by comprehensively processing multi-dimensional data of patients, and to issue early warning signals in a timely manner to support more timely and personalized treatment decision-making processes.

[0003] The Chinese patent with the announcement number CN115762788A discloses a tumor recurrence and metastasis risk assessment system based on tumor cell detection, including: a front-end device layer for obtaining basic data for tumor recurrence and metastasis assessment, a device control module for controlling the operation of the terminal associated with the front-end device layer, sending and receiving data processing instructions; and an analysis and evaluation module for identifying and processing basic data for tumor recurrence and metastasis assessment. This patent solves the problem that the prediction indicators of existing methods do not include the detection and analysis of tumor cells of postoperative patients, which affects the accuracy of the prediction, in order to improve the accuracy of tumor recurrence and metastasis risk assessment.

[0004] However, the patent has the following main problems: First, it lacks big data support. Although the system can collect and analyze the data of a single patient, it does not mention how to integrate a large amount of historical data to improve the accuracy of the model. Second, the personalized prediction ability is limited. The existing scheme relies on preset tumor recurrence rules and may not fully consider individual differences. Third, real-time and continuous monitoring are insufficient. The system focuses more on data analysis after regular examinations, and does not mention how to achieve long-term tracking and real-time warning of patients. Summary of the invention

[0005] The purpose of the present invention is to provide a tumor recurrence risk assessment and early warning system based on data analysis, which enhances personalized prediction capabilities by introducing big data support, and continuously monitors the patient's postoperative physical condition in real time by setting up a real-time monitoring interface in conjunction with a variety of medical devices, which can significantly improve the system's prediction accuracy and practicality, and solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A tumor recurrence risk assessment and early warning system based on data analysis, including:

[0008] A multi-data collection module is used to automatically obtain and integrate multi-dimensional data such as patients' personal information, medical records, laboratory test results, imaging reports, and biomarkers from multiple sources;

[0009] Big data processing platform, integrating massive historical case databases, and storing, cutting and preprocessing data;

[0010] The model architecture system includes a feature engineering module and a predictive modeling unit. The feature engineering module uses statistical methods and machine learning algorithms to extract key features and supports dynamic updating of feature libraries. The predictive modeling unit uses a deep neural network architecture (such as a convolutional neural network (CNN) or a recurrent neural network (RNN)) to build a personalized recurrence risk prediction model.

[0011] The individual data processing system includes: a real-time monitoring interface that connects to various smart medical devices to collect changes in patients' daily physiological parameters; a risk assessment engine that comprehensively analyzes current data and historical patterns to output the probability of recurrence and its confidence interval for each patient; a dynamic early warning mechanism that immediately sends notifications to doctors and / or patients when a high-risk state is detected; and a visual user interface that provides interactive charts to display assessment results and recommended measures.

[0012] Preferably, the big data processing platform includes:

[0013] Data storage unit, using distributed file systems such as HDFS or Amazon S3 to store large data sets. The system can divide data into multiple blocks and store them on different nodes of the cluster, thereby improving data reading speed and fault tolerance;

[0014] The data preprocessing unit uses ETL tools to clean, transform and load raw data, and uses the outlier detection component to identify and clean up inaccurate or abnormal data entries, including but not limited to removing duplicate records, filling missing values, and standardizing field formats.

[0015] Preferably, the feature engineering module is further used for:

[0016] Convert data of different scales to the same scale and convert categorical variables into numerical form using one-hot encoding;

[0017] Use several statistical methods and machine learning algorithms to select relevant features and remove insignificant features;

[0018] By using automated tools to automatically discover new feature combinations and establish a feature library management system, the results and performance indicators of each feature extraction are recorded, and the feature extraction process is run regularly for evaluation and update.

[0019] Preferably, the predictive modeling unit comprises a transfer learning strategy for quickly adapting to the data characteristics of new types of cancer, and the transfer learning strategy is further used to:

[0020] The model weights trained based on conventional tumor types are used as initial weights, and the learning rate is adjusted according to the amount of data;

[0021] Freeze the low-level feature extraction layers of the tumor pre-trained model, add new fully connected layers to focus on the specific features of the new task, compile the new tumor model and train it on the dataset of the new task;

[0022] Build a multi-task model with a shared base layer, specify the loss function and optimizer for each task, prepare the data of multiple tasks and combine them into a dictionary, and finally train the tumor model simultaneously on the datasets of multiple tasks.

[0023] Preferably, the real-time monitoring interface is further used to: unify the data structure using the FHIR format; support multiple wireless communication protocols such as Bluetooth and Wi-Fi to ensure the security and stability of data transmission; process data from different devices and support large-scale concurrent access.

[0024] Preferably, the intelligent medical device comprises:

[0025] Wearable devices, including smart bracelets / watches and health monitoring patches. Smart bracelets / watches are used to monitor patients' heart rate, blood oxygen saturation, sleep quality, activity level and body temperature. Health monitoring patches are used to monitor heart electrical activity, respiratory rate and changes in body position.

[0026] Home medical devices, including sphygmomanometers for monitoring blood pressure and pulse, glucometers for monitoring blood sugar levels, and weight scales for monitoring weight and body fat percentage;

[0027] Remote monitoring equipment, including pulmonary function testers for monitoring patients’ maximum expiratory volume and ease of breathing, and portable ultrasound devices for monitoring changes in tumor lesions;

[0028] A mobile app to record patients’ eating habits, post-operative feelings, and discomfort symptoms;

[0029] Environmental sensors are used to monitor pollutants such as PM2.5 and CO2 in indoor air and record environmental conditions that affect patient comfort and health.

[0030] Preferably, the risk assessment engine is further used to:

[0031] Feed the newly input data into the trained model to calculate the probability of recurrence;

[0032] Based on the uncertainty of the model, the confidence interval of the probability of recurrence was calculated and possible risk factors were analyzed;

[0033] The calculation and analysis results are presented to doctors and patients through a visual user interface, and explainable results are provided to help doctors understand the reasons for model decisions. When the calculation results show high risk, the dynamic early warning mechanism is automatically triggered and emergency notifications are sent to doctors and patients.

[0034] Automatically adjust weight distribution based on the latest medical research results to maintain the effectiveness of the model.

[0035] Preferably, the visual user interface provides multiple language versions to facilitate use by users who speak different languages.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention integrates large-scale historical case data and uses these rich data sets for model training, which can significantly improve the accuracy and generalization ability of the prediction model. By analyzing a large amount of patient data, the model can capture more subtle risk factors, thereby providing a more reliable recurrence risk assessment. Combined with the statistical methods and automatic feature selection technology in the feature engineering module, the prediction accuracy of the model is further improved.

[0038] 2. The present invention not only considers clinical pathological data, but also integrates multiple data sources such as genomic information, living habits, environmental factors, etc. to provide personalized risk assessment for each patient. Multi-dimensional data fusion enables the model to have a more comprehensive understanding of the patient's health status, thereby providing more accurate prediction results, and supports dynamic updating of the feature library. It can continuously optimize the feature combination according to the latest medical research results and technological advances, ensuring that the model always uses the latest and most representative features, further enhancing the ability of personalized prediction.

[0039] 3. The present invention connects various intelligent medical devices through a real-time monitoring interface to collect changes in patients' daily physiological parameters. It can continuously monitor key indicators such as heart rate, blood pressure, blood sugar, sleep quality, etc., and detect abnormal conditions in a timely manner. When the system detects a high-risk state, it will immediately send an alarm notification to the doctor and / or patient. This instant warning function helps to intervene early, take timely measures to prevent tumor recurrence, and improve treatment effects. Through continuous data collection and analysis, the system can dynamically adjust risk assessment results and provide patients with continuous health management advice. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0041] Figure 2It is a flow chart of the system operation of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0043] To address the problems of existing technologies lacking big data support, limited individual prediction capabilities, and insufficient real-time and continuous monitoring, please refer to Figure 1-2 , this embodiment provides the following technical solutions:

[0044] A tumor recurrence risk assessment and early warning system based on data analysis includes multiple data collection modules, a big data processing platform, a model architecture system and an individual data processing system.

[0045] The multi-data collection module is used to automatically acquire and integrate multi-dimensional data such as patients' personal information, medical records, laboratory test results, imaging reports, and biomarkers from multiple sources.

[0046] The data input layer is as follows:

[0047] Patient personal information: age, gender, weight, blood type, etc.

[0048] Clinical pathological data: tumor size, location, grade, stage, surgical method, postoperative physical therapy plan, etc.

[0049] Biomarker data: specific protein levels in the blood, gene expression data, etc.

[0050] Imaging data: CT scan, MRI, PET-CT and other image data.

[0051] Historical data: past medical history, family medical history, lifestyle, etc.

[0052] The big data processing platform integrates the above-mentioned past case data into a massive historical case database, and stores, cuts and pre-processes the data. The big data processing platform includes a data storage unit and a data pre-processing unit: the data storage unit uses a distributed file system such as HDFS or Amazon S3 to store large-scale data sets. The system can divide the data into multiple blocks and store them on different nodes of the cluster, thereby improving data reading speed and fault tolerance; the data pre-processing unit uses ETL tools to clean, transform and load the original data, and uses the outlier detection component to identify and clean up inaccurate or abnormal data entries, including but not limited to removing duplicate records, filling missing values, and standardizing field formats.

[0053] The model architecture system includes a feature engineering module and a predictive modeling unit. The feature engineering module uses statistical methods and machine learning algorithms to extract key features and supports dynamic updating of the feature library. The predictive modeling unit uses a deep neural network architecture (such as a convolutional neural network (CNN) or a recurrent neural network (RNN)) to build a personalized recurrence risk prediction model.

[0054] Among them, the feature engineering module is further used to:

[0055] Convert data of different scales to the same scale, and use one-hot encoding to convert categorical variables into numerical form; use several statistical methods and machine learning algorithms to select relevant features and eliminate insignificant features; use automated tools to automatically discover new feature combinations, establish a feature library management system, record the results and performance indicators of each feature extraction, and regularly run the feature extraction process for evaluation and update.

[0056] The example code of the feature engineering module is as follows:

[0057]

[0058]

[0059]

[0060]

[0061]

[0062] Through the above steps, the feature engineering module can effectively extract key features from the raw data and support dynamic updating of the feature library, thereby continuously optimizing the performance of the tumor recurrence risk assessment and early warning system.

[0063] The predictive modeling unit includes a transfer learning strategy for quickly adapting to the data characteristics of new types of cancer. The transfer learning strategy is further used to: use the model weights trained on the basis of conventional tumor types as the initial weights, and adjust the learning rate according to the amount of data; freeze the low-level feature extraction layer of the tumor pre-trained model, add a new fully connected layer, focus on the specific features of the new task, compile the new tumor model and train it on the data set of the new task; build a multi-task model with a shared base layer, specify the loss function and optimizer for each task, prepare the data of multiple tasks, and combine them into a dictionary, and finally train the tumor model on the data sets of multiple tasks at the same time.

[0064] The example code for running the prediction modeling unit is as follows:

[0065] from tensorflow.keras.models import load_model

[0066] Loading a pre-trained model

[0067] pre_trained_model=load_model('path / to / pretrained_model.h5')

[0068] Building a new model

[0069] new_model=Sequential()

[0070] for layer in pre_trained_model.layers:

[0071] new_model.add(layer)

[0072] Fine-tuning the model

[0073] new_model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])

[0074] new_model.fit(new_cancer_data,new_cancer_labels,epochs=10,batch_size=32,validation_split=0.2)

[0075] As shown above, when the amount of data for a new tumor type is small, the common features learned by the pre-trained model can be used to accelerate the convergence of the new model and reduce the risk of overfitting.

[0076] Loading VGG16 pre-trained model

[0077] base_model=VGG16(weights='imagenet',include_top=False,input_shape=(224,224,3))

[0078] Freeze Base Layer

[0079] for layer in base_model.layers:

[0080] layer.trainable = False

[0081] Adding a new fully connected layer

[0082] from tensorflow.keras.layers import Flatten,Dense

[0083] model=Sequential()

[0084] model.add(base_model)

[0085] model.add(Flatten())

[0086] model.add(Dense(256,activation='relu'))

[0087] model.add(Dense(1,activation='sigmoid'))

[0088] Compile a new model and train it on the new task

[0089] model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])

[0090] model.fit(new_cancer_images,new_cancer_labels,epochs=10,batch_size=32,validation_split=0.2)

[0091] As shown above, the fixed low-level feature extraction layers can provide stable feature representations, and the new fully connected layers can focus on learning specific features for new tasks and improve the accuracy of the model.

[0092] Construct a multi-task model with a shared base layer, where each task has its own output layer as follows:

[0093] from tensorflow.keras.layers import Input,Conv2D,MaxPooling2D,Flatten,Dense

[0094] Input Layer

[0095] input_layer=Input(shape=(224,224,3))

[0096] Shared Convolutional Layers

[0097] x=Conv2D(32,(3,3),activation='relu')(input_layer)

[0098] x = MaxPooling2D((2,2))(x)

[0099] x=Conv2D(64,(3,3),activation='relu')(x)

[0100] x = MaxPooling2D((2,2))(x)

[0101] Branch 1: Task 1

[0102] branch1=Flatten()(x)

[0103] branch1=Dense(128,activation='relu')(branch1)

[0104] output1=Dense(1,activation='sigmoid',name='task1')(branch1)

[0105] Branch 2: Task 2

[0106] branch2=Flatten()(x)

[0107] branch2=Dense(128,activation='relu')(branch2)

[0108] output2=Dense(1,activation='sigmoid',name='task2')(branch2)

[0109] Building a multi-task model

[0110] multi_task_model=Model(inputs=input_layer,outputs=[output1,output2])

[0111] Specify loss function and optimizer for each task

[0112] multi_task_model.compile(optimizer='adam',

[0113] loss={'task1':'binary_crossentropy','task2':'binary_crossentropy'},

[0114] metrics={'task1':'accuracy','task2':'accuracy'})

[0115] Prepare multiple task data and combine them into a dictionary

[0116] data={'input_1':combined_images,

[0117] 'task1':task1_labels,

[0118] 'task2':task2_labels}

[0119] Training models on multiple task datasets simultaneously

[0120] multi_task_model.fit(data,epochs=10,batch_size=32,validation_split=0.2)

[0121] As shown above, by sharing the underlying features, the generalization ability of the model can be improved and it helps to handle tasks with smaller amounts of data.

[0122] The individual data processing system includes a real-time monitoring interface, a risk assessment engine, a dynamic early warning mechanism, and a visual user interface, as follows:

[0123] The real-time monitoring interface connects various smart medical devices to collect changes in patients' daily physiological parameters and unifies the data structure using the FHIR format; it supports multiple wireless communication protocols such as Bluetooth and Wi-Fi to ensure the security and stability of data transmission; it processes data from different devices and supports large-scale concurrent access.

[0124] The smart medical devices connected to the real-time monitoring interface include but are not limited to:

[0125] Wearable devices, including smart bracelets / watches and health monitoring patches. Smart bracelets / watches are used to monitor patients' heart rate, blood oxygen saturation, sleep quality, activity level and body temperature. Health monitoring patches are used to monitor heart electrical activity, respiratory rate and changes in body position.

[0126] Home medical devices, including sphygmomanometers for monitoring blood pressure and pulse, glucometers for monitoring blood sugar levels, and weight scales for monitoring weight and body fat percentage;

[0127] Remote monitoring equipment, including pulmonary function testers for monitoring patients’ maximum expiratory volume and ease of breathing, and portable ultrasound devices for monitoring changes in tumor lesions;

[0128] A mobile app to record patients’ eating habits, post-operative feelings, and discomfort symptoms;

[0129] Environmental sensors are used to monitor pollutants such as PM2.5 and CO2 in indoor air and record environmental conditions that affect patient comfort and health.

[0130] The risk assessment engine comprehensively analyzes current data and historical patterns, and outputs the probability of recurrence and its confidence interval for each patient. Specifically, the newly input data is fed into the trained model to calculate the probability of recurrence; based on the uncertainty of the model, the confidence interval of the probability of recurrence is calculated, and possible risk factors are analyzed; the calculation and analysis results are presented to doctors and patients through a visual user interface, and explainable results are provided to help doctors understand the reasons for the model's decision. When the calculation results show a high risk, the dynamic early warning mechanism is automatically triggered and an emergency notification is sent to doctors and patients; the weight distribution is automatically adjusted according to the latest medical research results to maintain the effectiveness of the model.

[0131] Dynamic early warning mechanism, when a high-risk state is detected, immediately sends a notification to the doctor and / or patient to improve responsiveness.

[0132] The visual user interface provides interactive charts to display assessment results and recommended measures. It is available in multiple languages ​​to facilitate users who speak different languages. By providing an intuitive and easy-to-use operation interface, it is convenient for medical staff to view analysis results and manage patient files. At the same time, patients can also obtain their own health data and risk assessment reports through mobile applications, etc., to increase their understanding and participation in their own health.

[0133] Working principle: The system first automatically collects and integrates multi-dimensional data such as patients' personal information, medical records, laboratory test results, imaging reports, and biomarkers from multiple sources, and continuously collects daily physiological parameters through smart medical devices. Next, data preprocessing is performed, including data cleaning, standardization, and coding. The feature engineering module uses statistical methods and machine learning algorithms to extract key features and supports dynamic updating of feature libraries. Advanced machine learning algorithms are used in the model training phase, combined with transfer learning strategies to quickly adapt to the data characteristics of new types of cancer. The trained model is used to calculate the probability of recurrence and confidence interval for each patient, and an immediate warning mechanism is used to send alerts to doctors and patients when high risks are detected. The system also provides a visual user interface to display analysis results and personalized health recommendations, support long-term tracking and management, and improve treatment compliance and patient engagement. Through this process, the system can achieve efficient and accurate risk assessment and early warning, and improve patients' survival rate and quality of life.

[0134] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0135] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A tumor recurrence risk assessment and early warning system based on data analysis, characterized in that: include: A multi-data collection module is used to automatically obtain and integrate multi-dimensional data such as patients' personal information, medical records, laboratory test results, imaging reports, and biomarkers from multiple sources; Big data processing platform, integrating massive historical case databases, and storing, cutting and preprocessing data; Model architecture system, including feature engineering module and predictive modeling unit. The feature engineering module uses statistical methods and machine learning algorithms to extract key features and supports dynamic update of feature library. The predictive modeling unit uses a deep neural network architecture to build a personalized recurrence risk prediction model; The individual data processing system includes: a real-time monitoring interface that connects to various smart medical devices to collect changes in patients' daily physiological parameters; a risk assessment engine that comprehensively analyzes current data and historical patterns to output the probability of recurrence and its confidence interval for each patient; a dynamic early warning mechanism that immediately sends notifications to doctors and / or patients when a high-risk state is detected; and a visual user interface that provides interactive charts to display assessment results and recommended measures.

2. A tumor recurrence risk assessment and early warning system based on data analysis according to claim 1, characterized in that: The big data processing platform includes: Data storage unit, which uses distributed file systems such as HDFS or Amazon S3 to store large data sets. The system divides the data into multiple blocks and stores them on different nodes of the cluster to improve data reading speed and fault tolerance; The data preprocessing unit uses ETL tools to clean, transform and load raw data, and uses the outlier detection component to identify and clean up inaccurate or abnormal data entries, including but not limited to removing duplicate records, filling missing values, and standardizing field formats.

3. A tumor recurrence risk assessment and early warning system based on data analysis according to claim 2, characterized in that: The feature engineering module is further used to: Convert data of different scales to the same scale and convert categorical variables into numerical form using one-hot encoding; Use several statistical methods and machine learning algorithms to select relevant features and remove insignificant features; By using automated tools to automatically discover new feature combinations and establish a feature library management system, the results and performance indicators of each feature extraction are recorded, and the feature extraction process is run regularly for evaluation and update.

4. The tumor recurrence risk assessment and early warning system based on data analysis according to claim 3, characterized in that: The predictive modeling unit includes a transfer learning strategy for quickly adapting to the data characteristics of new types of cancer, and the transfer learning strategy is further used to: The model weights trained based on conventional tumor types are used as initial weights, and the learning rate is adjusted according to the amount of data; Freeze the low-level feature extraction layers of the tumor pre-trained model, add new fully connected layers to focus on the specific features of the new task, compile the new tumor model and train it on the dataset of the new task; Build a multi-task model with a shared base layer, specify the loss function and optimizer for each task, prepare the data of multiple tasks and combine them into a dictionary, and finally train the tumor model simultaneously on the datasets of multiple tasks.

5. A tumor recurrence risk assessment and early warning system based on data analysis according to claim 4, characterized in that: The real-time monitoring interface is further used to: unify the data structure using the FHIR format; support multiple wireless communication protocols such as Bluetooth and Wi-Fi to ensure the security and stability of data transmission; process data from different devices and support large-scale concurrent access.

6. A tumor recurrence risk assessment and early warning system based on data analysis according to claim 5, characterized in that: The intelligent medical device comprises: Wearable devices, including smart bracelets / watches and health monitoring patches. Smart bracelets / watches are used to monitor patients' heart rate, blood oxygen saturation, sleep quality, activity level and body temperature. Health monitoring patches are used to monitor heart electrical activity, respiratory rate and changes in body position. Home medical devices, including sphygmomanometers for monitoring blood pressure and pulse, glucometers for monitoring blood sugar levels, and weight scales for monitoring weight and body fat percentage; Remote monitoring equipment, including pulmonary function testers for monitoring patients’ maximum expiratory volume and ease of breathing, and portable ultrasound devices for monitoring changes in tumor lesions; A mobile app to record patients’ eating habits, post-operative feelings, and discomfort symptoms; Environmental sensors are used to monitor pollutants such as PM2.5 and CO2 in indoor air and record environmental conditions that affect patient comfort and health.

7. A tumor recurrence risk assessment and early warning system based on data analysis according to claim 6, characterized in that: The risk assessment engine is further configured to: Feed the newly input data into the trained model to calculate the probability of recurrence; Based on the uncertainty of the model, the confidence interval of the probability of recurrence was calculated and possible risk factors were analyzed; The calculation and analysis results are presented to doctors and patients through a visual user interface, and explainable results are provided to help doctors understand the reasons for model decisions. When the calculation results show high risk, the dynamic early warning mechanism is automatically triggered and emergency notifications are sent to doctors and patients. Automatically adjust weight distribution based on the latest medical research results to maintain the effectiveness of the model.

8. The tumor recurrence risk assessment and early warning system based on data analysis according to claim 7, characterized in that: The visual user interface provides multiple language versions to facilitate use by users who speak different languages.

Images