Diagnostic system and device for ileocecum ulcer and storage medium

Abstract

The invention discloses an ileocecum ulcer differential diagnosis system and device and a storage medium, and the system comprises a prediction diagnosis module which is configured to execute the following operations: based on received first disease diagnosis information corresponding to a target object, extracting a plurality of pieces of first feature information; preprocessing the plurality of pieces of first feature information to generate a plurality of pieces of second feature information, and combining the plurality of pieces of second feature information to determine one or more pieces of first combined information; determining a first prediction and diagnosis model corresponding to each piece of second feature information, and performing prediction by using the first prediction and diagnosis model and the corresponding second feature information so as to generate a first prediction result; determining a second prediction diagnosis model corresponding to the one or more pieces of first combined information, and performing prediction by using the second prediction diagnosis model and the corresponding combined information so as to generate a second prediction result; and fusing the first prediction result and the second prediction result, and generating prediction diagnosis information.

Description

Technical Field

[0001] The present application relates to the field of medical automated diagnosis technology, and in particular to a differential diagnosis system, device and storage medium for ileocecal ulcer. Background Art

[0002] Ileocecal ulcer is one of the common disease manifestations of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. Medical personnel can make a differential diagnosis of ileocecal ulcer to distinguish whether the patient is suffering from Crohn's disease, intestinal tuberculosis or intestinal Behcet's disease. However, ileocecal ulcer, a common pathological manifestation of the digestive system, is still a major diagnostic problem. Due to the high incidence of intestinal tuberculosis, intestinal tuberculosis has become one of the main causes of ileocecal ulcer. In addition, in recent years, the incidence of Crohn's disease and intestinal Behcet's disease has also shown an increasing trend year by year, which further increases the difficulty of differential diagnosis of ileocecal ulcer.

[0003] In this context, the medical community has been committed to developing new differential diagnostic tools and methods to improve the accuracy of differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. In recent years, with the continuous development of artificial intelligence and deep learning technology, deep learning models based on visual convolutional neural network feature maps have shown good potential in helping to differentially diagnose Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease.

[0004] However, there are still some defects in the above-mentioned use of deep learning models to differentially diagnose Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. For example, since the existing deep learning model only relies on the patient's endoscopic images (i.e., colonoscopic images) to differentially diagnose Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, it only considers the information of a single modality of endoscopic images, ignoring other clinical data and imaging data that may be valuable for differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. As a result, for endoscopic images of atypical lesions, the accuracy of differential diagnosis is low, and it is unable to provide accurate and more comprehensive decision support for doctors.

[0005] With regard to the technical problem that the prior art only considers information from a single modality and ignores other clinical data and imaging data that may be valuable for the differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, the accuracy of differential diagnosis for endoscopic images of atypical lesions is low, and it is unable to provide doctors with accurate and more comprehensive decision support, no effective solution has been proposed so far. Summary of the invention

[0006] The embodiments of the present disclosure provide a differential diagnosis system, device and storage medium for ileocecal ulcer, so as to at least solve the technical problem existing in the prior art that the prior art only considers information of a single modality and ignores other clinical data and imaging data that may be valuable for the differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. As a result, for endoscopic images of atypical lesions, the accuracy of differential diagnosis is low, and it is impossible to provide doctors with accurate and more comprehensive decision support.

[0007] According to one aspect of an embodiment of the present disclosure, a differential diagnosis system for ileocecal ulcer is provided, comprising: a database, a predictive diagnosis module and an output module, wherein the database is connected to the predictive diagnosis module and configured to provide first disease diagnosis information corresponding to a target object; the output module is connected to the predictive diagnosis module and configured to output the predictive diagnosis information of the target object, wherein the predictive diagnosis information is used to indicate that the target object suffers from Crohn's disease, intestinal tuberculosis or intestinal Behcet's disease, and wherein the predictive diagnosis module is configured to perform the following operations: extract multiple first feature information based on the received first disease diagnosis information corresponding to the target object; preprocess the multiple first feature information to generate multiple second feature information, and combine each second feature information to determine one or more first combination information; determine a first predictive diagnosis model corresponding to each second feature information, and use the first predictive diagnosis model and the corresponding second feature information for prediction, thereby generating a first prediction result; determine a second predictive diagnosis model corresponding to one or more first combination information, and use the second predictive diagnosis model and the corresponding combination information for prediction, thereby generating a second prediction result; and fuse the first prediction result and the second prediction result to generate predictive diagnosis information.

[0008] According to another aspect of an embodiment of the present disclosure, a storage medium is further provided, the storage medium including a stored program, wherein when the program is running, a processor executes any one of the methods described above.

[0009] According to another aspect of the embodiments of the present disclosure, a differential diagnosis device for ileocecal ulcer is also provided, including: a feature information extraction module, used to extract multiple first feature information based on the received first disease diagnosis information corresponding to the target object; a feature information combination module, used to preprocess the multiple first feature information, generate multiple second feature information, and combine each second feature information to determine one or more first combination information; a first prediction result generation module, used to determine the first prediction diagnosis model corresponding to each second feature information, and use the first prediction diagnosis model and the corresponding second feature information to predict, so as to generate a first prediction result; a second prediction result generation module, used to determine the second prediction diagnosis model corresponding to one or more first combination information, and use the second prediction diagnosis model and the corresponding combination information to predict, so as to generate a second prediction result; and a prediction result fusion module, used to fuse the first prediction result and the second prediction result, and generate predictive diagnosis information.

[0010] According to another aspect of the embodiments of the present disclosure, a differential diagnosis device for ileocecal ulcer is also provided, including: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: extracting multiple first feature information based on the received first disease diagnosis information corresponding to the target object; preprocessing the multiple first feature information to generate multiple second feature information, and combining each second feature information to determine one or more first combination information; determining a first predictive diagnosis model corresponding to each second feature information, and using the first predictive diagnosis model and the corresponding second feature information for prediction, thereby generating a first prediction result; determining a second predictive diagnosis model corresponding to one or more first combination information, and using the second predictive diagnosis model and the corresponding combination information for prediction, thereby generating a second prediction result; and fusing the first prediction result and the second prediction result to generate predictive diagnosis information.

[0011] The present application discloses a differential diagnosis system for ileocecal ulcer. The system includes a database, a predictive diagnosis module and an output module. The database is connected to the predictive diagnosis module and configured to provide first disease diagnosis information corresponding to a target object. The output module is connected to the predictive diagnosis module and configured to output the predictive diagnosis information of the target object. The predictive diagnosis module can also perform the following operations: first, based on the received first disease diagnosis information corresponding to the target object, a plurality of first feature information is extracted. Then, the plurality of first feature information is preprocessed to generate a plurality of second feature information, and each second feature information is combined to determine one or more first combination information. Further, a first predictive diagnosis model corresponding to each feature information is determined, and the first predictive diagnosis model and the corresponding feature information are used to make a prediction, thereby generating a first prediction result. Then, a second predictive diagnosis model corresponding to one or more first combination information is determined, and the second predictive diagnosis model and the corresponding combination information are used to make a prediction, thereby generating a second prediction result. Finally, the first prediction result and the second prediction result are fused, and the predictive diagnosis information is generated.

[0012] Since the predictive diagnosis module in the present application not only generates the corresponding first prediction results based on multiple first feature information, but also further generates one or more first combination information based on multiple first feature information, and generates the corresponding second prediction results based on one or more first combination information, the predictive diagnosis information corresponding to the target object determined based on the first prediction result and the second prediction result is the differential diagnosis information determined based on multi-modal feature information. Therefore, compared with the existing differential diagnosis information determined based on feature information of a single modality, the present application can achieve the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support.

[0013] This solves the technical problem that the prior art only considers information from a single modality and ignores other clinical data and imaging data that may be valuable for the differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. As a result, the accuracy of differential diagnosis for endoscopic images of atypical lesions is low, and it is unable to provide doctors with accurate and more comprehensive decision support.

[0014] In addition, since the present application further performs a fusion process on the first prediction result and the second prediction result after generating the first prediction result and the second prediction result, the accuracy of the predictive diagnosis information generated after the fusion can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present application. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0016] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to Embodiment 1 of the present disclosure;

[0017] Figure 2 is a schematic diagram of the differential diagnosis system for ileocecal ulcer according to Example 1 of the present application;

[0018] Figure 3 is a schematic diagram of a predictive diagnosis module according to Example 1 of the present application;

[0019] Figure 4 is a flowchart of a method for performing operations according to the predictive diagnosis module described in Example 1 of the present application;

[0020] Figure 5 is a schematic diagram of a feature selection interface according to Example 1 of the present application;

[0021] Figure 6 is a schematic diagram of a feature filling interface according to Example 1 of the present application;

[0022] Figure 7 is a schematic diagram of the predictive diagnosis result interface according to Example 1 of the present application;

[0023] Figure 8 It is a schematic diagram of the disease error rate corresponding to the third predictive diagnosis model in the training phase according to Example 1 of the present application, and the disease error rate corresponding to the first predictive diagnosis model and the second predictive diagnosis model in the application phase;

[0024] Fig. 9 is a schematic diagram of the model combination method according to Example 1 of the present application;

[0025] Fig.10 is a schematic diagram of the differential diagnosis device for ileocecal ulcer according to Example 2 of the present application; and

[0026] Fig.11 This is a schematic diagram of the differential diagnosis device for ileocecal ulcer according to Example 3 of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0029] Example 1

[0030] According to this embodiment, a method embodiment of differential diagnosis of ileocecal ulcer is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0031] The method embodiment provided in this embodiment can be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computing device for implementing a differential diagnosis method for ileocecal ulcer. Figure 1 As shown, the computing device may include one or more processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, the transmission device, and the input / output interface are connected to the processor via a bus. In addition, it may also include: a display, a keyboard, and a cursor control device connected to the input / output interface. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0032] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computing device. As involved in the embodiments of the present disclosure, the data processing circuitry acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0033] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the differential diagnosis method of ileocecal ulcer in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the differential diagnosis method of ileocecal ulcer of the above-mentioned application program is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0034] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computing device. In one example, the transmission device includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.

[0035] The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of the computing device.

[0036] It should be noted that, in some optional embodiments, the above Figure 1 The computing device shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computing devices described above.

[0037] Figure 2 Schematic diagram of the differential diagnosis system for ileocecal ulcer according to this embodiment. Figure 2 As shown, the system includes: a database, a predictive diagnosis module and an output module. The predictive diagnosis module is connected to the database and is used to obtain first disease diagnosis information corresponding to the target object from the database. The first disease diagnosis information includes, for example, clinical information extracted from an electronic medical record information system, endoscopic information extracted from an endoscopic report, and CT information extracted from a CT report.

[0038] In addition, the predictive diagnosis module is also connected to the output module, and is used to send the predictive diagnosis information of the target object to the terminal device connected to the peripheral device through the output module. The predictive diagnosis information is used to indicate that the target object suffers from Crohn's disease, intestinal tuberculosis or intestinal Behcet's disease.

[0039] Figure 3 is a schematic diagram of a predictive diagnosis module according to an embodiment of the present application. Figure 3 As shown, the prediction diagnosis module includes: a feature extraction module, a preprocessing module, a model prediction module and a fusion module. The feature extraction module is connected to the preprocessing module, and the feature extraction module is used to extract multiple first feature information from the first disease diagnosis information. The preprocessing module is used to normalize and encode the multiple first feature information and generate multiple second feature information. In addition, the preprocessing module is also used to combine the respective second feature information and determine one or more first combination information.

[0040] The model prediction module is connected to the preprocessing module and is used to use the first prediction diagnosis model corresponding to each second feature information to make a prediction and generate a first prediction result. In addition, the model prediction module is also used to use the second prediction diagnosis model corresponding to one or more first combination information to make a prediction and generate a second prediction result.

[0041] The fusion module is connected to the model prediction module, and the fusion module is used to fuse the first prediction result and the second prediction result to generate prediction diagnosis information.

[0042] It should be noted that the database, predictive diagnosis module and output module in the system can all be applied to the hardware structure described above.

[0043] Under the above operating environment, according to the first aspect of this embodiment, a differential diagnosis system for ileocecal ulcer is provided. Figure 4 A flowchart showing the corresponding operations performed by the predictive diagnosis module is shown. Figure 4 As shown, the predictive diagnosis module is also configured to perform the following operations:

[0044] S402: extracting a plurality of first feature information based on the received first disease diagnosis information corresponding to the target object;

[0045] S404: Preprocessing the plurality of first feature information to generate a plurality of second feature information, and combining the second feature information to determine one or more first combination information;

[0046] S406: Determine a first prediction and diagnosis model corresponding to each piece of second feature information, and perform prediction using the first prediction and diagnosis model and the corresponding second feature information, thereby generating a first prediction result;

[0047] S408: Determine a second prediction and diagnosis model corresponding to one or more first combination information, and use the second prediction and diagnosis model and the corresponding combination information to perform prediction, thereby generating a second prediction result; and

[0048] S410: The first prediction result and the second prediction result are integrated to generate prediction diagnosis information.

[0049] Specifically, refer to Figure 2 and Figure 3 As shown, the model prediction module is connected to the database, so that the feature extraction module can obtain the first disease diagnosis information corresponding to the target object from the database through the data transmission interface. The first disease diagnosis information includes, for example, clinical information extracted from the electronic medical record information system, endoscopic image information extracted from the endoscopic report, and CT image information extracted from the CT report.

[0050] Afterwards, the feature extraction module extracts multiple first feature information based on the received first disease diagnosis information corresponding to the target object (S402). For example, the feature extraction module can extract clinical features from clinical information using natural language processing, can also extract endoscopic features from endoscopic image information using natural language processing, and can also extract CT features from CT image information using natural language processing. In addition, it is worth noting that the target object's medical records may not necessarily have all three types of information (i.e., clinical information, endoscopic image information, and CT image information). For example, some target objects' medical records only have clinical information, some target objects' medical records only have endoscopic image information, and some target objects' medical records only have clinical information and endoscopic image information. That is, there may be only one or more first feature information corresponding to the target object, which will not be repeated here.

[0051] In addition, in order to ensure the integrity and consistency of the data, each first feature information is organized into a form of tabular data for subsequent data processing and model training, thereby achieving the technical effect of maximizing the accuracy and robustness of the model (i.e., the first predictive diagnosis model and the second predictive diagnosis model).

[0052] It is worth noting that the multiple first feature information extracted by the feature extraction module will be displayed in the data entry front-end interface. And for the features that the feature extraction module cannot automatically capture, the feature extraction module will display missing information on the data entry front-end interface and prompt the target object to enter manually. For example, because the feature extraction module cannot extract clinical features from the first disease diagnosis information, the data entry front-end interface will display missing clinical information and prompt the target object to manually enter the clinical information.

[0053] Furthermore, in the case where the target object cannot manually enter the missing information, the predictive diagnosis module will provide default values ​​(i.e., the mean value for continuous features and the mode value for categorical features) based on historical training data to fill in the missing values, thereby ensuring that the subsequent operation can proceed normally. The historical training data is stored in a database, for example.

[0054] For example, when there is missing data in the first disease diagnosis information, the missing rate of each first characteristic information is calculated. If the missing rate of any first characteristic information among multiple first characteristic information exceeds 20%, it is considered that the information contained in the first characteristic information is insufficient and may not provide a useful basis for judgment for the first predictive diagnosis model and the second predictive diagnosis model. Therefore, it is decided to delete the first characteristic information with a higher missing rate. Therefore, when deleting the first characteristic information with a higher missing rate, the remaining first characteristic information with a lower missing rate is retained to ensure the quality of the data and the performance of the prediction model.

[0055] In the case where the feature extraction module extracts multiple first feature information corresponding to the target object, the preprocessing module preprocesses the multiple first feature information to generate multiple second feature information, and combines each second feature information to determine one or more first combination information (S404). Specifically, the preprocessing module can, for example, save multiple first feature information in a server, and perform preprocessing operations (i.e., normalization processing and one-hot encoding processing) on ​​each first feature information to generate multiple second feature information. This further determines that the multiple second feature information is in the best state before being input into the first predictive diagnosis model, avoiding model misjudgment due to inconsistent data format or range.

[0056] For example, for continuous variables, the preprocessing module normalizes multiple first feature information and scales the variable values ​​of the first feature information to the interval of [0, 1] to eliminate the dimensional differences between the first feature information and avoid excessive influence of a certain first feature information on the first predictive diagnosis model and the second predictive diagnosis model. For multi-classification variables, a one-hot encoding method is used to represent each category as a binary variable, so that the first predictive diagnosis model and the second predictive diagnosis model can better process the category information. In addition, one-hot encoding of multiple first feature information can eliminate the order information between the categorical variables and avoid misjudgment of the first predictive diagnosis model and the second predictive diagnosis model due to the order relationship between the categories.

[0057] Further, after the preprocessing module preprocesses the plurality of first feature information to generate the plurality of second feature information, the preprocessing module combines the respective second feature information to determine one or more first combination information. For example, the second feature information after processing includes clinical features, endoscopic features, and CT features, so that the preprocessing module can combine the respective second feature information with each other and generate the first combination information of clinical features + endoscopic features, the first combination information of endoscopic features + CT features, the first combination information of clinical features + CT features, and the first combination information of clinical features + endoscopic features + CT features.

[0058] In addition, it is worth noting that the examples provided above include multiple second feature information (i.e., clinical features, endoscopic features, and CT features), but based on the above content, it can be seen that since the first disease diagnosis information may have missing information, it may also include only one second feature information or two second feature information. Therefore, in the case of only one second feature information, there is no first combination information. In the case of including two second feature information, only one corresponding first combination information is included.

[0059] Afterwards, the model prediction module determines the first predictive diagnosis model corresponding to each second feature information, and uses the first predictive diagnosis model and the corresponding second feature information for prediction, thereby generating a first prediction result (S406). Specifically, the model prediction module inputs each second feature information into each first predictive diagnosis model stored in the server, and each first predictive diagnosis model predicts based on the corresponding second feature information. For example, the second feature information includes clinical features, endoscopic features, and CT features. And after the first predictive diagnosis model training is completed, each second feature information has been evaluated in advance, and the model with the best performance for each second feature information is selected as the final corresponding first predictive diagnosis model.

[0060] For example, when the second feature information is a clinical feature, the random forest classifier performs best (i.e., the first predictive diagnostic model corresponding to the clinical feature), then the random forest classifier is used to make predictions based on the clinical feature, thereby generating a first prediction result corresponding to the clinical feature.

[0061] When the second feature information is an endoscopic feature, the extreme gradient boosting classifier performs best (i.e., the first predictive diagnostic model corresponding to the endoscopic feature), and the extreme gradient boosting classifier is used to perform prediction based on the endoscopic feature to generate a first prediction result corresponding to the endoscopic feature.

[0062] When the second feature information is a CT feature, the lightweight gradient boosting classifier performs best (i.e., the first predictive diagnostic model corresponding to the CT feature), and the lightweight gradient boosting classifier is used to perform prediction based on the CT feature to generate a first prediction result corresponding to the CT feature.

[0063] Therefore, the model prediction module can predict the corresponding second characteristic information based on the first prediction diagnosis model corresponding to each second characteristic information, thereby generating a first prediction result. And because the first prediction diagnosis model corresponding to each second characteristic information is a predetermined prediction diagnosis model that performs best for each second characteristic information, each first prediction diagnosis model can quickly process the input second characteristic information and output the first prediction result, thereby providing auxiliary decision support for medical personnel.

[0064] Similarly, the model prediction module determines the second predictive diagnosis model corresponding to each first combination of information, and uses the second predictive diagnosis model and the corresponding first combination of information for prediction, thereby generating a second prediction result (S408). Specifically, the model prediction module inputs each first combination into each second predictive diagnosis model stored in the server, and each second predictive diagnosis model predicts according to the corresponding first combination of information. For example, the first combination of information includes clinical features + endoscopic features, endoscopic features + CT features, clinical features + CT features, and clinical features + endoscopic features + CT features. And after the training of the second predictive diagnosis model is completed, each first combination of information has been evaluated in advance, and the model with the best performance for each first combination of information has been selected as the final corresponding second predictive diagnosis model.

[0065] For example, when the first combination information is clinical characteristics + endoscopic characteristics, the extreme random tree classifier performs best (i.e., the second predictive diagnostic model corresponding to clinical characteristics + endoscopic characteristics), then the extreme random tree classifier is used to make predictions based on the clinical characteristics + endoscopic characteristics, thereby generating a second prediction result corresponding to the clinical characteristics + endoscopic characteristics.

[0066] When the first combined information is clinical characteristics + CT features, the extreme random tree classifier performs best (i.e., the second predictive diagnostic model corresponding to clinical characteristics + CT features), then the extreme random tree classifier is used to make predictions based on the clinical characteristics + CT features, thereby generating a second prediction result corresponding to the clinical characteristics + CT features.

[0067] When the first combined information is endoscopic features + CT features, the extreme gradient boosting classifier performs best (i.e., the second predictive diagnostic model corresponding to the endoscopic features + CT features), then the extreme gradient boosting classifier is used to perform prediction based on the endoscopic features + CT features, thereby generating a second prediction result corresponding to the endoscopic features + CT features.

[0068] When the first combination information is clinical characteristics + endoscopic characteristics + CT features, the lightweight gradient boosting classifier performs best (i.e., the second predictive diagnostic model corresponding to clinical characteristics + endoscopic characteristics + CT features), then the lightweight gradient boosting classifier is used to make predictions based on the clinical characteristics + endoscopic characteristics + CT features, thereby generating a second prediction result corresponding to the clinical characteristics + endoscopic characteristics + CT features.

[0069] Therefore, the model prediction module can predict the corresponding first combination information based on the second prediction diagnosis model corresponding to each first combination information, thereby generating a second prediction result. And because the second prediction diagnosis model corresponding to each first combination information is a predetermined prediction diagnosis model that performs best for each first combination information, each second prediction diagnosis model can quickly process the input first combination information and output the second prediction result, thereby providing auxiliary decision support for medical personnel.

[0070] In addition, it is worth noting that the above-described example is a case where multiple second feature information is included, and the second prediction result is generated based on the second prediction diagnosis model corresponding to the multiple first combination information. It is worth noting that those skilled in the art that when two second feature information are included, the specific implementation method is the same as above, so it will not be repeated here.

[0071] Finally, the fusion module fuses the first prediction result and the second prediction result to generate predictive diagnosis information corresponding to the target object (S410). For example, if only two second feature information are included, it means that two first prediction results and one second prediction result will eventually be generated, so the fusion module fuses the two second feature information and one second prediction result, and generates predictive diagnosis information corresponding to the target object. For another example, if three second feature information are included, it means that three first prediction results and four second prediction results will eventually be generated, so the fusion module fuses the three first prediction results and the four second prediction results, and generates predictive diagnosis information corresponding to the target object. The above content will be described in detail later, so it will not be repeated here.

[0072] Thus, when the fusion module generates predictive diagnostic information corresponding to the target object, the predictive diagnostic information is further transmitted to the output interface of the terminal device connected to the peripheral device through the output module. Among them, the output interface can not only display the probability distribution of the above three diseases (i.e., Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease), but also give the final diagnosis suggestions to help medical personnel make more accurate diagnosis and treatment.

[0073] Figure 5 It is a schematic diagram of a feature selection interface according to an embodiment of the present application. Figure 6 It is a schematic diagram of a feature filling interface according to an embodiment of the present application. Figure 7 is a schematic diagram of a prediction diagnosis result interface according to an embodiment of the present application. Figure 5 , Figure 6 and Figure 7 As shown, first, medical personnel can view an intuitive and easy-to-use feature selection interface through the interface displayed by the terminal device, thereby ensuring that medical personnel can select the required feature information. For example, medical personnel can select clinical features and endoscopic features as feature information used by the system.

[0074] Secondly, medical personnel can also fill in characteristic factors, so that the system can further obtain valuable information. For example, after clicking "Endoscopic Features", medical personnel can choose to fill in characteristic information related to endoscopic features on the next interface. These include the number of ulcers, the maximum ulcer diameter, whether there are deep ulcers, whether there are longitudinal ulcers, whether there are annular ulcers, and whether there are irregular ulcers.

[0075] Finally, medical personnel can, for example, view the predictive diagnosis information displayed in the form of an Echart chart through a terminal device connected to an external device, thereby further improving the readability of the predictive diagnosis information. Among them, the predictive diagnosis information may, for example, include the risk probability of the above three diseases suffered by the target object (i.e., Crohn's disease, intestinal tuberculosis or intestinal Behcet's disease) and the recommended diagnosis conclusion. For example, after the medical personnel fill in the above characteristic factors, the system automatically generates predictive diagnosis information corresponding to the target object and displays it on the terminal device. Among them, the predictive diagnosis information shows that the probability of the target object suffering from Crohn's disease is greater than the probability of suffering from intestinal tuberculosis and the probability of suffering from intestinal Behcet's disease, and further displays the diagnosis conclusion as "the patient is recommended to be diagnosed with Crohn's disease."

[0076] Therefore, through the system provided in this application, medical personnel can more efficiently and accurately identify Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, thereby greatly improving the reliability of differential diagnosis and the timeliness of treatment.

[0077] As described in the background technology, the existing use of deep learning models for differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease still has some defects. For example, since the existing deep learning model only relies on the patient's endoscopic images (i.e., colonoscopic images) to differentially diagnose Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, it only considers the information of a single modality of endoscopic images, ignoring other clinical data and imaging data that may be valuable for differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. As a result, for endoscopic images of atypical lesions, the accuracy of differential diagnosis is low, and it is unable to provide accurate and more comprehensive decision support for doctors.

[0078] In view of this, the present application provides a differential diagnosis system for ileocecal ulcers. And because the predictive diagnosis module in the present application not only generates a corresponding first prediction result based on multiple first feature information, but also further generates one or more first combination information based on multiple first feature information, and generates a corresponding second prediction result based on one or more first combination information, the predictive diagnosis information corresponding to the target object determined based on the first prediction result and the second prediction result is differential diagnosis information determined based on multi-modal feature information. Therefore, compared with the existing differential diagnosis information determined based on feature information of a single modality, the present application can achieve the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support.

[0079] And because this application constructs a diagnostic framework that integrates clinical features, endoscopic features, and CT features, it is possible to more comprehensively reflect the characteristics and pathological manifestations of the disease by integrating these three different medical information sources, thereby achieving a more comprehensive disease classification and diagnosis. This multi-dimensional information combined diagnostic strategy has significant advantages in the differential diagnosis of complex diseases.

[0080] This solves the technical problem that the prior art only considers information from a single modality and ignores other clinical data and imaging data that may be valuable for the differential diagnosis of Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. As a result, the accuracy of differential diagnosis for endoscopic images of atypical lesions is low, and it is unable to provide doctors with accurate and more comprehensive decision support.

[0081] Optionally, the predictive diagnosis module is also configured to perform the following operations: when a second feature information is extracted, determine the corresponding first predictive diagnosis model; use the first predictive diagnosis model and the corresponding second feature information to perform prediction, thereby generating a third prediction result; and use the third prediction result as predictive diagnosis information.

[0082] Specifically, based on the above-mentioned contents, it can be known that when the feature extraction module extracts only one first feature information according to the first disease diagnosis information corresponding to the target object, the preprocessing module only has one second feature information after preprocessing.

[0083] Thus, when there is only one second feature information, a first prediction diagnosis model corresponding to the second feature information is determined. For example, when the second feature information only includes clinical features, a random forest classifier is used to perform a prediction diagnosis based on the clinical features, thereby generating a third prediction result corresponding to the clinical features.

[0084] For another example, when the second feature information is an endoscopic feature, an extreme gradient boosting classifier is used to perform prediction based on the endoscopic feature, thereby generating a third prediction result corresponding to the endoscopic feature.

[0085] For another example, when the second feature information is a CT feature, a lightweight gradient boosting classifier is used to perform prediction based on the CT feature, thereby generating a third prediction result corresponding to the CT feature.

[0086] Therefore, when there is only one second feature information and only the third prediction result is generated, the fusion module directly uses the third prediction result as the prediction diagnosis information and displays the prediction diagnosis information on the terminal device through the output module.

[0087] Thus, through the above operations, the technical effect of ensuring the output of predictive diagnosis information corresponding to the target object and enabling medical personnel to better provide treatment decisions for the target object based on the predictive diagnosis information is achieved.

[0088] Optionally, the predictive diagnosis module includes: a preprocessing module, wherein the preprocessing module is configured to perform the following operations: normalize and encode the multiple first feature information, and generate corresponding multiple second feature information.

[0089] Specifically, after the feature extraction module generates multiple first feature information based on the first disease diagnosis information corresponding to the target object, the preprocessing module can normalize and one-hot code the multiple first feature information. For example, for continuous variables, the preprocessing module normalizes the multiple first feature information and scales the variable value of the first feature information to the interval of [0, 1] to eliminate the dimensional differences between the various first feature information and avoid the influence of a certain first feature information on the model. For multi-classification variables, the one-hot code encoding method is used to represent each category as a binary variable, so that the first predictive diagnosis model and the second predictive diagnosis model can better process the category information. In addition, the one-hot code encoding of multiple first feature information can eliminate the order information between the classification variables and avoid the first predictive diagnosis model and the second predictive diagnosis model from making misjudgments due to the order relationship between the categories.

[0090] Therefore, the preprocessing module is used to normalize the multiple first feature information and perform one-hot encoding processing, thereby achieving the technical effect of ensuring that the data is in the best state before being input into the first predictive diagnosis model and the second predictive diagnosis model, and avoiding model misjudgment due to inconsistent data format or range.

[0091] Optionally, the operation of fusing the first prediction result and the second prediction result to generate predictive diagnostic information includes: determining the error rate of each third predictive diagnostic model in predicting Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, wherein each third predictive diagnostic model includes the first predictive diagnostic model and the second predictive diagnostic model; based on the error rate, determining the total error rate of multiple model combinations corresponding to the first predictive diagnostic model and the second predictive diagnostic model, and based on the total error rate, determining the average error rate corresponding to different model combinations; selecting the minimum average error rate, and using the model combination corresponding to the minimum average error rate as the target model combination; and fusing the first prediction result and the second prediction result corresponding to the target model combination to generate predictive diagnostic information.

[0092] Specifically, before using the first predictive diagnosis model corresponding to each second feature information and the second predictive diagnosis model corresponding to multiple first combination information, it is necessary to pre-train the first predictive diagnosis model and the second predictive diagnosis model, and determine the optimal model corresponding to each feature information and the optimal model corresponding to each combination information. In this embodiment, for example, based on the accuracy of the prediction results of each third predictive diagnosis model for different feature information and combination information, the fourth predictive diagnosis model corresponding to each feature information and each combination information can be determined. Among them, the fourth predictive diagnosis model is used to indicate the optimal model corresponding to each feature information and combination information.

[0093] Figure 8 Schematic diagram of the disease error rate corresponding to the third predictive diagnosis model in the training phase and the disease error rate corresponding to the first predictive diagnosis model and the second predictive diagnosis model in the application phase according to the embodiment of the present application. Figure 8 As shown, in the model training stage, the evaluation module determines the error rate of the third predictive diagnosis model 1 to n in predicting Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease. For example, the clinical feature sample 1 corresponding to the second sample object 1 is respectively input into multiple third predictive diagnosis models 1 to n, and the predictive diagnosis results corresponding to each third predictive diagnosis model 1 to n are obtained. Thus, when the true diagnosis result corresponding to the second sample object 1 is determined, the error rate corresponding to each third predictive diagnosis model for the clinical feature sample can be determined based on the true diagnosis result and the predictive diagnosis result.

[0094] Similarly, the clinical feature samples 2 corresponding to the second sample object 2 are respectively input into a plurality of third prediction diagnosis models 1 to n, and the prediction diagnosis results corresponding to the respective third prediction diagnosis models 1 to n are obtained. Thus, when the true diagnosis result corresponding to the second sample object 2 is determined, the error rate corresponding to each third prediction diagnosis model for the clinical feature sample can be determined based on the true diagnosis result and the prediction diagnosis result.

[0095] And so on.

[0096] The clinical feature samples x corresponding to the second sample object x are respectively input into a plurality of third predictive diagnosis models 1 to n, and the predictive diagnosis results corresponding to the respective third predictive diagnosis models 1 to n are obtained. Thus, when the true diagnosis result corresponding to the second sample object x is determined, the error rate corresponding to each third predictive diagnosis model for the clinical feature samples can be determined based on the true diagnosis result and the predictive diagnosis result.

[0097] Therefore, based on the above operations, the evaluation module can determine the error rate of each third prediction diagnosis model 1 for the clinical feature samples 1-x corresponding to each second sample object 1-x.

[0098] Referring to the above content, when determining the predictive diagnosis results of the second predictive diagnosis model corresponding to other feature information samples and combined information samples of the second sample objects 1~x, the error rates corresponding to each third predictive diagnosis model can be further determined based on the actual diagnosis results and the predicted diagnosis results.

[0099] Based on the above steps, the error rates of the three different diseases predicted and diagnosed by the third prediction and diagnosis models can be determined. Figure 8 For the third predictive diagnostic model 1) shown, when the true diagnostic results and predicted diagnostic results of the second sample objects 1~x are determined, the error rate of predictive diagnosis for Crohn's disease using the random forest classifier (i.e., the third predictive diagnostic model 1) can be further determined; the error rate of predictive diagnosis for intestinal tuberculosis using the random forest classifier (i.e., the third predictive diagnostic model 1) can be further determined; and the error rate of predictive diagnosis for intestinal Behcet's disease using the random forest classifier (i.e., the third predictive diagnostic model 1) can be further determined.

[0100] Similarly, the error rates corresponding to the third predictive diagnosis models 2-n predicting and diagnosing Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease can be determined.

[0101] Furthermore, when the evaluation module determines the error rate of predicting Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease corresponding to each third predictive diagnostic model, the fusion module randomly combines each first predictive diagnostic model corresponding to each second feature information and each second predictive diagnostic model corresponding to each first combination information based on each second feature information and each first combination information, and determines multiple model combination methods.

[0102] For example, the target object includes three second feature information (i.e., clinical features, endoscopic features, and CT features), and four first combination information (i.e., clinical features + endoscopic features, endoscopic features + CT features, clinical features + CT features, and clinical features + endoscopic features + CT features). And the model prediction module has input each second feature information into the corresponding first predictive diagnosis model, and obtained the corresponding first prediction result. The model prediction module has input each first combination information into the corresponding second predictive diagnosis model, and obtained the corresponding second prediction result. Thus, the fusion module can further determine the error rate corresponding to the first prediction result based on the first prediction result output by the first predictive diagnosis model and the error rate of predicting three diseases corresponding to each third predictive diagnosis model 1 to n determined in the training phase.

[0103] refer to Figure 8As shown, for example, after the model prediction module inputs the clinical features (i.e., the second feature information) into the random forest classifier (i.e., the first prediction diagnosis model 1), it predicts and diagnoses that the target object may suffer from intestinal tuberculosis (i.e., the first prediction result). The random forest classifier corresponds to the third prediction diagnosis model 1 in the training stage. Thus, the fusion module can further determine the error rate of the random forest classifier (i.e., the first prediction diagnosis model 1) in predicting intestinal tuberculosis (i.e., the first prediction result) in the application stage based on the intestinal tuberculosis corresponding to the random forest classifier (i.e., the first prediction diagnosis model 1) and the error rate determined in the training stage when the random forest classifier (i.e., the third prediction diagnosis model 1) predicts intestinal tuberculosis.

[0104] Based on the same operations as above, the fusion module can determine the error rates corresponding to the first diagnostic results predicted by each first predictive diagnostic model 2 to 3 and the error rates corresponding to the second diagnostic results predicted by each second predictive diagnostic model 1 to 4 during the application phase.

[0105] Furthermore, the fusion module randomly combines the first predictive diagnosis models corresponding to the clinical features, endoscopic features, and CT features, respectively, and the second predictive diagnosis models corresponding to the clinical features + endoscopic features, endoscopic features + CT features, clinical features + CT features, and clinical features + endoscopic features + CT features, and determines 2 7 A combination of models.

[0106] Fig. 9 Schematic diagram of the model combination method according to the embodiment of the present application. Fig. 9 As shown, for example, multiple model combinations include: 1. First predictive diagnostic model 1 + first predictive diagnostic model 2 + second predictive diagnostic model 1. 2. First predictive diagnostic model 1 + first predictive diagnostic model 2 + first predictive diagnostic model 3 + second predictive diagnostic model 1 + second predictive diagnostic model 2 + second predictive diagnostic model 3 + second predictive diagnostic model 4. 3. First predictive diagnostic model 1 + first predictive diagnostic model 2 + first predictive diagnostic model 3 + second predictive diagnostic model 1 + second predictive diagnostic model 2 + second predictive diagnostic model 3, and so on.

[0107] When the fusion module determines multiple model combinations of the first predictive diagnosis model and the second predictive diagnosis model, the total error rate corresponding to the multiple model combinations can be further determined. For example, since the fusion module has determined the error rates corresponding to each first predictive diagnosis model and the error rates corresponding to each second predictive diagnosis model, when the fusion module determines multiple model combinations, the total error rate corresponding to each model combination can be determined. Fig. 9As shown, for example, one of the model combinations is: first predictive diagnostic model 1, first predictive diagnostic model 2, and second predictive diagnostic model 1. Thus, the fusion module adds the error rate corresponding to the first prediction result predicted by the first predictive diagnostic model 1, the error rate corresponding to the first prediction result predicted by the first predictive diagnostic model 2, and the error rate corresponding to the second prediction result predicted by the second predictive diagnostic model 1, and determines the total error rate corresponding to the model combination.

[0108] Similarly, based on the same method as above, the fusion module can determine the total error rate corresponding to the combination with other models.

[0109] Furthermore, the fusion module can determine the average error rate corresponding to each model combination after determining the total error rate corresponding to each model combination. For example, if the total error rate corresponding to the model combination of the first predictive diagnosis model 1, the first predictive diagnosis model 2, and the second predictive diagnosis model 1 is 10%, then the average error rate corresponding to the model combination is 3.333%.

[0110] After that, the fusion module selects the minimum average error rate after determining the average error rates corresponding to each model combination, and determines the model combination corresponding to the minimum average error rate as the target model combination. For example, the average error rate corresponding to the model combination of the first predictive diagnosis model 1, the first predictive diagnosis model 2, and the second predictive diagnosis model 1 is 3.333%. The average error rate corresponding to the model combination of the first predictive diagnosis model 1 + the first predictive diagnosis model 2 + the first predictive diagnosis model 3 + the second predictive diagnosis model 1 + the second predictive diagnosis model 2 + the second predictive diagnosis model 3 + the second predictive diagnosis model 4 is 5%. This means that the average error rate corresponding to the model combination of the first predictive diagnosis model 1, the first predictive diagnosis model 2, and the second predictive diagnosis model 1 is the smallest, and the model combination of the first predictive diagnosis model 1, the first predictive diagnosis model 2, and the second predictive diagnosis model 1 is the target model combination.

[0111] Finally, the fusion module fuses the first prediction result and the second prediction result corresponding to the target model combination mode, and generates predictive diagnosis information. For example, first, when the first predictive diagnosis model 1 outputs the corresponding feature vector based on clinical features, the first predictive diagnosis model 2 outputs the corresponding feature vector based on endoscopic features, and the second predictive diagnosis model 2 outputs the corresponding feature vector based on clinical features + endoscopic features, the fusion module determines the inverse error rate corresponding to the first predictive diagnosis model 1, the inverse error rate corresponding to the first predictive diagnosis model 2, and the inverse error rate corresponding to the second predictive diagnosis model 2, and then the fusion module determines the proportion of the inverse error rate of the first predictive diagnosis model 1 to the total error rate, determines the proportion of the inverse error rate of the first predictive diagnosis model 2 to the total error rate, and determines the proportion of the inverse error rate of the second predictive diagnosis model 1 to the total error rate.

[0112] Then, the fusion module uses the above-determined proportional relationship corresponding to the first prediction diagnosis model 1, the first prediction diagnosis model 2, and the second prediction diagnosis model 1 as the corresponding weight value, and uses the determined weight value to weight the corresponding first prediction result and the second prediction result to generate the prediction diagnosis information. For example, the weight value w1 corresponding to the first prediction diagnosis model 1, the weight value w2 corresponding to the first prediction diagnosis model 2, and the weight value w3 corresponding to the second prediction diagnosis model 1 can determine the final prediction result vector based on the following formula:

[0113] I s =I1×w1+I2×w2+I3×w3

[0114] Among them, I s represents the final prediction vector after weighted summation, I1 represents the feature vector corresponding to the first prediction result output by the first prediction diagnosis model 1, I2 represents the feature vector corresponding to the first prediction result output by the first prediction diagnosis model 2, and I3 represents the feature vector corresponding to the second prediction result output by the second prediction diagnosis model 1.

[0115] Finally, the fusion module generates corresponding prediction diagnosis information based on the final prediction result vector. The fusion module can, for example, input the prediction result vector into a pre-set softmax classifier to output probability values ​​corresponding to Crohn's disease, intestinal tuberculosis, and intestinal Behcet's disease.

[0116] Therefore, through the above-mentioned operation of fusing the first prediction result corresponding to the first predictive diagnosis model and the second prediction result corresponding to the second predictive diagnosis model, and generating predictive diagnosis information, the technical effect of outputting more accurate diagnostic information is achieved, thereby helping medical personnel to make better decisions.

[0117] Optionally, the preprocessing module is also configured to perform the following operations: obtain multiple first feature information samples corresponding to each first sample object from the database, and perform normalization and encoding processing on the multiple first feature information samples, thereby generating multiple second feature information samples corresponding to each first sample object. Further optionally, the predictive diagnosis module also includes: a model training module, wherein the model training module is communicatively connected to the preprocessing module and configured to perform the following operations: receive multiple second feature information samples corresponding to each first sample object sent by the preprocessing module, and combine each second feature information sample to determine multiple second combination information corresponding to each first sample object; and construct multiple third predictive diagnosis models, and use the multiple second feature information samples corresponding to each first sample object and the corresponding multiple second combination information to train the multiple third predictive diagnosis models respectively.

[0118] Specifically, before using the first predictive diagnosis model to generate the first prediction result based on the plurality of second feature information predictions, and before using the second predictive diagnosis model to generate the second prediction result based on the plurality of first combination information predictions, the first predictive diagnosis model and the second predictive diagnosis model need to be trained in advance. In addition, the database stores, for example, a plurality of first feature information samples corresponding to a plurality of first sample objects, and a plurality of second feature information samples corresponding to a plurality of second sample objects, and the plurality of first feature information samples and the plurality of second feature information samples are feature information extracted and stored in the database via a feature extraction module. It is worth noting that the plurality of first feature information samples corresponding to the plurality of first sample objects are used to train the plurality of third predictive diagnosis models, and the plurality of third feature information samples corresponding to the plurality of second sample objects are used to detect the performance of the plurality of third predictive diagnosis models. In addition, in order to ensure the generalization ability of the plurality of third predictive diagnosis models, the plurality of first sample objects and the plurality of second sample objects can be randomly divided in a ratio of 8:2. That is, 80% of the data is used to train the third predictive diagnosis model, and 20% of the data is used to test the performance of the third predictive diagnosis model.

[0119] In addition, after the data samples are divided, multiple first feature information samples corresponding to multiple first sample objects and multiple third feature information samples corresponding to multiple second sample objects are further interpolated for data. The interpolation method adopts a multiple interpolation method. The multiple interpolation method generates multiple possible values ​​through multiple interpolations to reflect the uncertainty of the data, thereby significantly improving the robustness and prediction accuracy of the first predictive diagnosis model and the second predictive diagnosis model. The multiple interpolation method is not only applicable to the interpolation of continuous variables, but also performs well for the interpolation of categorical variables.

[0120] Thus, firstly, the preprocessing module is also connected to the database for communication, and the database stores a plurality of first feature information samples corresponding to each first sample object. Thus, the preprocessing module can obtain a plurality of first feature information samples corresponding to each first sample object from the database. Then, the preprocessing module performs normalization processing and one-hot encoding processing on the obtained plurality of first feature information samples, thereby generating a plurality of second feature information samples corresponding to each first sample object. For example, the first sample object 1 includes 3 first feature information samples, the first sample object 2 includes 2 first feature information samples, ..., the first sample object m includes 3 first feature information samples. Thus, the preprocessing module performs normalization processing and one-hot encoding processing on the plurality of first feature information samples corresponding to each first sample object 1 to m, respectively, and generates a plurality of second feature information samples corresponding to each first sample object 1 to m.

[0121] In addition, after the preprocessing module uses normalization and one-hot encoding to convert the first feature information sample into the second feature information sample, the Boruta algorithm is further used to select the second feature information sample. Among them, the Boruta algorithm is a feature selection method based on random forests, which can effectively identify features that contribute to the model and eliminate redundant or irrelevant features. Through the Boruta algorithm, feature selection can be performed on clinical feature samples, endoscopic feature samples, and CT feature samples (ie, second feature information samples) respectively, so as to determine that the first predictive diagnosis model and the second predictive diagnosis model can obtain the best performance when trained using the most relevant feature samples.

[0122] Afterwards, the model training module receives multiple second feature information samples corresponding to each first sample object sent by the preprocessing module, and respectively combines the multiple second feature information samples corresponding to each first sample object, thereby determining multiple second combination information corresponding to each first sample object. For example, the first sample object 1 corresponds to 3 first feature information samples, namely, a clinical feature sample, an endoscopic feature sample, and a CT feature sample. Thus, the model training module can combine the above 3 first feature information samples to generate multiple second combination information (i.e., clinical feature sample + endoscopic feature sample, clinical feature sample + CT feature sample, endoscopic feature sample + CT feature sample, and clinical feature sample + endoscopic feature sample + CT feature sample).

[0123] For another example, the first sample object 2 corresponds to two first feature information samples, namely a clinical feature sample and an endoscopic feature sample. Thus, the model training module can combine the above two first feature information samples to generate a second combination information (ie, clinical feature sample + endoscopic feature sample).

[0124] Therefore, it can be known from the above that the plurality of first sample objects 1-m respectively correspond to one or more pieces of second combination information.

[0125] Furthermore, the model training module constructs multiple third predictive diagnosis models. The multiple third predictive diagnosis models constructed by the model training module include, but are not limited to, a logistic regression model, a gradient boosting classifier, a K nearest neighbor classifier, a random forest classifier, an extreme gradient boosting classifier, a lightweight gradient boosting classifier, an extreme random tree classifier, a support vector machine classifier, and a naive Bayes classifier.

[0126] Then, the model training module trains multiple third predictive diagnosis models using multiple second feature information samples corresponding to each first sample object and one or more second combination information. For example, the first sample object 1 corresponds to 3 first feature information samples (i.e., clinical feature samples, endoscopic feature samples, and CT feature samples) and 4 second combination information (i.e., clinical feature samples + endoscopic feature samples, clinical feature samples + CT feature samples, endoscopic feature samples + CT feature samples, and clinical feature samples + endoscopic feature samples + CT feature samples).

[0127] The first sample object 2 corresponds to two first feature information samples (ie, a clinical feature sample and an endoscopic feature sample) and one second combination information (ie, a clinical feature sample+endoscopic feature sample).

[0128] And so on.

[0129] The first sample object n corresponds to 3 first feature information samples (i.e., clinical feature sample, endoscopic feature sample and CT feature sample) and 4 second combination information (i.e., clinical feature sample + endoscopic feature sample, clinical feature sample + CT feature sample, endoscopic feature sample + CT feature sample and clinical feature sample + endoscopic feature sample + CT feature sample).

[0130] Thus, the model training module inputs the multiple clinical feature samples 1-a corresponding to the respective first sample objects 1-m into multiple third prediction and diagnosis models, and trains the multiple third prediction and diagnosis models respectively, wherein a≤m.

[0131] Similarly, the model training module inputs a plurality of endoscopic feature samples 1-b corresponding to each of the first sample objects 1-m into a plurality of third prediction and diagnosis models, and trains the plurality of third prediction and diagnosis models respectively.

[0132] The model training module inputs a plurality of CT feature samples 1-c corresponding to each of the first sample objects 1-m into a plurality of third prediction and diagnosis models, and trains the plurality of third prediction and diagnosis models respectively, wherein c≤m.

[0133] The model training module inputs a plurality of clinical feature samples + endoscopic feature samples 1-d corresponding to each of the first sample objects 1-m into a plurality of third predictive diagnosis models, and trains the plurality of third predictive diagnosis models respectively, wherein d≤m.

[0134] The model training module inputs a plurality of clinical feature samples + CT feature samples 1 to e corresponding to each of the first sample objects 1 to m into a plurality of third prediction and diagnosis models, and trains the plurality of third prediction and diagnosis models respectively, wherein e≤m.

[0135] The model training module inputs a plurality of endoscopic feature samples + CT feature samples 1-f corresponding to each of the first sample objects 1-m into a plurality of third predictive diagnosis models, and trains the plurality of third predictive diagnosis models respectively, wherein f≤m.

[0136] And so on.

[0137] The model training module inputs a plurality of clinical feature samples + endoscopic feature samples + CT feature samples 1-g corresponding to each first sample object 1-m into a plurality of third prediction diagnosis models, and trains the plurality of third prediction diagnosis models respectively, wherein g≤m.

[0138] The above operations thus provide the necessary basis for subsequently using the first predictive diagnostic model and the second predictive diagnostic model to determine the first prediction result and the second prediction result corresponding to the target object, thereby ensuring that medical personnel can use the system provided by the present application to quickly determine the technical effect of the disease suffered by the target object.

[0139] It is worth noting that in the process of training the first predictive diagnosis model and the second predictive diagnosis model, the first feature information samples corresponding to each first sample object are used to optimize the parameters of the first predictive diagnosis model and the second predictive diagnosis model, and hyperparameter tuning is performed through 10-fold cross-validation and grid search. The 10-fold cross-validation can effectively evaluate the generalization ability of the first predictive diagnosis model and the second predictive diagnosis model to avoid overfitting problems, while the grid search is used to find the best hyperparameter combination to improve the performance of the first predictive diagnosis model and the second predictive diagnosis model.

[0140] In addition, the multiple third predictive diagnosis models mentioned above cover multiple types of models such as linear models, decision tree models, integrated models and probability models, which can provide different perspectives to deal with complex relationships in the data, thereby improving the accuracy of the output predictive diagnosis information.

[0141] In order to ensure that the first prediction diagnosis model and the second prediction diagnosis model can be reused in practical applications, the present application uses Python's pickle library to save the optimal model corresponding to each feature information as a .pkl file. The above optimal model is securely stored in the server for subsequent prediction tasks and classification system construction.

[0142] Optionally, the preprocessing module is also configured to perform the following operations: obtain multiple third feature information samples corresponding to each second sample object from the database, and perform normalization and encoding processing on the multiple third feature information samples, thereby generating multiple fourth feature information samples corresponding to each second sample object. Further optionally, the predictive diagnosis module also includes: an evaluation module, wherein the evaluation module is connected to the model training module, and the evaluation module is configured to perform the following operations: receive multiple fourth feature information samples corresponding to each second sample object sent by the preprocessing module, and combine each fourth feature information sample, thereby determining multiple third combination information corresponding to each second sample object; and using the multiple fourth feature information samples corresponding to each second sample object and the multiple third combination information corresponding to each second sample object, determine multiple fourth predictive diagnosis models corresponding to each fourth feature information sample and each third combination information, respectively, wherein the fourth predictive diagnosis model is used to indicate the optimal model corresponding to the fourth feature information sample and the third combination information.

[0143] Specifically, after the model training model trains multiple third predictive diagnosis models using multiple first feature information samples corresponding to each first sample object, the performance of the multiple third predictive diagnosis models is further tested, and the optimal predictive diagnosis model corresponding to each feature information is determined.

[0144] Thus, firstly, the preprocessing module can obtain multiple third feature information samples corresponding to each second sample object from the database, and then the preprocessing module normalizes and one-hot encodes the obtained multiple third feature information samples to generate multiple fourth feature information samples corresponding to each second sample object.

[0145] Then, the evaluation module receives the plurality of fourth feature information samples corresponding to the second sample objects sent by the preprocessing module, and respectively combines the plurality of fourth feature information samples corresponding to the second sample objects, thereby determining a plurality of third combination information corresponding to the second sample objects. That is, referring to the above example, it can be seen that the plurality of second sample objects respectively correspond to one or more third combination information.

[0146] Further, the evaluation module uses multiple fourth feature information samples and multiple third combination information to evaluate the trained multiple third prediction diagnosis models, so as to select the optimal prediction diagnosis model corresponding to each feature information. For example, the clinical feature samples 1 to r corresponding to the second sample objects 1 to x are input into the random forest classifier, and the corresponding prediction diagnosis results are obtained. The clinical feature samples 1 to r corresponding to the second sample objects 1 to x are input into the logistic regression model, and the corresponding prediction diagnosis results are obtained. The clinical feature samples 1 to r corresponding to the second sample objects 1 to x are input into the logistic gradient boosting classifier, and the corresponding prediction diagnosis results are obtained. The clinical feature samples 1 to r corresponding to the second sample objects 1 to x are input into the K nearest neighbor classifier, and the corresponding prediction diagnosis results are obtained. And so on. The clinical feature samples 1 to r corresponding to the second sample objects 1 to x are input into the naive Bayes classifier, and the corresponding prediction diagnosis results are obtained.

[0147] Then, the evaluation module determines the accuracy of the prediction results obtained by each of the third prediction diagnosis models according to each of the fourth feature information samples and each of the third combination information based on the actual diagnosis results corresponding to each of the second sample objects and the above-generated prediction diagnosis results. And when the accuracy of the prediction results corresponding to the random forest classifier is the highest, it means that the random forest classifier is the optimal prediction diagnosis model (i.e., the fourth prediction diagnosis model) for the clinical feature information.

[0148] Similarly, based on the same operations as above, the evaluation module can determine the optimal predictive diagnostic model corresponding to the endoscopic feature information, the optimal predictive diagnostic model corresponding to the CT feature information, the optimal predictive diagnostic model corresponding to the clinical feature information + endoscopic feature information, the optimal predictive diagnostic model corresponding to the clinical feature information + CT feature information, the optimal predictive diagnostic model corresponding to the endoscopic feature information + CT feature information, and the optimal predictive diagnostic model corresponding to the clinical feature information + endoscopic feature information + CT feature information.

[0149] Therefore, by utilizing the evaluation module to determine the operation of the optimal predictive diagnosis model corresponding to each feature information and each combination information, the accuracy of the predictive diagnosis information generated in subsequent predictive diagnosis tasks can be further improved, thereby providing a reference for the clinical diagnosis of medical personnel.

[0150] Therefore, according to the first aspect of this embodiment, the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support can be achieved.

[0151] The purpose of the present invention is to build a machine learning model that accurately distinguishes Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease by integrating multimodal information such as clinical symptoms, laboratory tests, imaging examinations and endoscopic examinations. Unlike existing models, the present invention uses machine learning technology rather than deep learning to model multimodal information including clinical information, endoscopic manifestations and CT data to improve the accuracy and applicability of differential diagnosis. The system provided by the present invention can not only perform well in typical cases, but also significantly improve the diagnostic accuracy of atypical cases, providing more comprehensive decision support for clinicians.

[0152] In addition, this application also carries out the following system design processing for the provided system:

[0153] Technical packaging of algorithm models: Through technical means, the algorithm model is technically packaged to determine the availability and stability of the algorithm model, which is also convenient for the technical foundation and maintenance of subsequent model updates. The specific technical implementation paths are as follows:

[0154] (1) Development environment configuration: Make sure the Python environment is installed, including the required joblib library and dependencies.

[0155] (2) Model loading: Implement a Python function or class to load the joblib model file and ensure that the model can be loaded into memory.

[0156] (3) API development: Use Python web frameworks such as Flask or Django to develop RESTful APIs, which will serve as interfaces for the front-end to call models.

[0157] (4) API request processing: Design the API endpoint to receive the full request, including the necessary input parameters and data.

[0158] (5) Data preprocessing: Convert the data sent by the front end into the format required by the model.

[0159] (6) Model prediction: Integrate model prediction logic in the API and use the loaded joblib model to predict the processed data.

[0160] (7) Result encapsulation: Encapsulate the model’s prediction results into JSON or other formats so that the front end can parse and display them.

[0161] (8) Performance optimization: Perform performance testing and optimization on the API as needed to ensure that the speed and processing power meet the requirements.

[0162] In addition, reference Figure 1As shown, according to the second aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0163] Therefore, according to this embodiment, the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support can be achieved.

[0164] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0165] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0166] Example 2

[0167] Fig.10 The differential diagnosis device 1000 for ileocecal ulcer according to this embodiment is shown, and the device 1000 corresponds to the method according to embodiment 1. Fig.10As shown, the device 1000 includes: a feature information extraction module 1010, which is used to extract multiple first feature information based on the received first disease diagnosis information corresponding to the target object; a feature information combination module 1020, which is used to preprocess the multiple first feature information, generate multiple second feature information, and combine each second feature information to determine one or more first combination information; a first prediction result generation module 1030, which is used to determine the first prediction diagnosis model corresponding to each second feature information, and use the first prediction diagnosis model and the corresponding second feature information to predict, so as to generate a first prediction result; a second prediction result generation module 1040, which is used to determine the second prediction diagnosis model corresponding to one or more first combination information, and use the second prediction diagnosis model and the corresponding combination information to predict, so as to generate a second prediction result; and a prediction result fusion module 1050, which is used to fuse the first prediction result and the second prediction result, and generate prediction diagnosis information.

[0168] Optionally, the device 1000 also includes: a predictive diagnosis model determination module, which is used to determine the corresponding first predictive diagnosis model when a second feature information is extracted; a third prediction result generation module, which is used to use the first predictive diagnosis model and the corresponding second feature information to make a prediction, thereby generating a third prediction result; and a predictive diagnosis information generation module, which is used to use the third prediction result as predictive diagnosis information.

[0169] Optionally, the feature information combination module 1020: a preprocessing module, used to perform normalization and encoding processing on multiple first feature information, and generate corresponding multiple second feature information.

[0170] Optionally, the prediction result fusion module 1050 includes: an error rate determination module, which is used to determine the error rate of each third prediction diagnostic model in predicting Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, wherein each third prediction diagnostic model includes a first prediction diagnostic model and a second prediction diagnostic model; an average error rate determination module, which is used to determine the total error rate of multiple model combinations corresponding to the first prediction diagnostic model and the second prediction diagnostic model based on the error rate, and determine the average error rate corresponding to different model combinations based on the total error rate; a target model combination determination module, which is used to select the minimum average error rate and use the model combination corresponding to the minimum average error rate as the target model combination; and a prediction result fusion submodule, which is used to fuse the first prediction result and the second prediction result corresponding to the target model combination, and generate predictive diagnosis information.

[0171] Optionally, the preprocessing module also includes: a first preprocessing submodule, which is used to obtain multiple first feature information samples corresponding to each first sample object from the database, and normalize and encode the multiple first feature information samples to generate multiple second feature information samples corresponding to each first sample object; and a second preprocessing submodule, which is used to obtain multiple third feature information samples corresponding to each second sample object from the database, and normalize and encode the multiple third feature information samples to generate multiple fourth feature information samples corresponding to each second sample object.

[0172] Optionally, the device 800 also includes: a first combination module, which is used to receive multiple second feature information samples corresponding to each first sample object sent by the preprocessing module, and combine each second feature information sample to determine multiple second combination information corresponding to each first sample object; and a model training module, which is used to construct multiple third predictive diagnosis models, and use the multiple second feature information samples corresponding to each first sample object, and the corresponding multiple second combination information to train the multiple third predictive diagnosis models respectively.

[0173] Optionally, the device 800 also includes: a second combination module, which is used to receive multiple fourth feature information samples corresponding to each second sample object sent by the preprocessing module, and combine each fourth feature information sample to determine multiple third combination information corresponding to each second sample object; and an evaluation module, which is used to use the multiple fourth feature information samples corresponding to each second sample object and the multiple third combination information corresponding to each second sample object to determine multiple fourth predictive diagnosis models corresponding to each fourth feature information sample and each third combination information, respectively, wherein the fourth predictive diagnosis model is used to indicate the optimal model corresponding to the fourth feature information sample and the third combination information.

[0174] Therefore, according to this embodiment, the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support can be achieved.

[0175] Example 3

[0176] Fig.11 A differential diagnosis device 1100 for ileocecal ulcer according to this embodiment is shown, and the device 1100 corresponds to the method according to embodiment 1. Fig.11As shown, the device 1100 includes: a processor 1110; and a memory 1120, which is connected to the processor 1110 and is used to provide the processor 1110 with instructions for processing the following processing steps: extracting multiple first feature information based on the received first disease diagnosis information corresponding to the target object; preprocessing the multiple first feature information to generate multiple second feature information, and combining each second feature information to determine one or more first combination information; determining a first prediction diagnosis model corresponding to each second feature information, and using the first prediction diagnosis model and the corresponding second feature information to predict, thereby generating a first prediction result; determining a second prediction diagnosis model corresponding to one or more first combination information, and using the second prediction diagnosis model and the corresponding combination information to predict, thereby generating a second prediction result; and fusing the first prediction result and the second prediction result to generate prediction diagnosis information

[0177] Optionally, the predictive diagnosis module is also configured to perform the following operations: when a second feature information is extracted, determine the corresponding first predictive diagnosis model; use the first predictive diagnosis model and the corresponding second feature information to perform prediction, thereby generating a third prediction result; and use the third prediction result as predictive diagnosis information.

[0178] Optionally, the predictive diagnosis module includes: a preprocessing module, wherein the preprocessing module is configured to perform the following operations: normalize and encode the multiple first feature information, and generate corresponding multiple second feature information.

[0179] Optionally, the operation of fusing the first prediction result and the second prediction result to generate predictive diagnostic information includes: determining the error rate of each third predictive diagnostic model in predicting Crohn's disease, intestinal tuberculosis and intestinal Behcet's disease, wherein each third predictive diagnostic model includes the first predictive diagnostic model and the second predictive diagnostic model; based on the error rate, determining the total error rate of multiple model combinations corresponding to the first predictive diagnostic model and the second predictive diagnostic model, and based on the total error rate, determining the average error rate corresponding to different model combinations; selecting the minimum average error rate, and using the model combination corresponding to the minimum average error rate as the target model combination; and fusing the first prediction result and the second prediction result corresponding to the target model combination to generate predictive diagnostic information.

[0180] Optionally, the preprocessing module is also configured to perform the following operations: obtaining multiple first feature information samples corresponding to each first sample object from the database, and normalizing and encoding the multiple first feature information samples to generate multiple second feature information samples corresponding to each first sample object; and obtaining multiple third feature information samples corresponding to each second sample object from the database, and normalizing and encoding the multiple third feature information samples to generate multiple fourth feature information samples corresponding to each second sample object.

[0181] Optionally, the predictive diagnosis module also includes: a model training module, wherein the model training module is communicatively connected to the preprocessing module and configured to perform the following operations: receiving multiple second feature information samples corresponding to each first sample object sent by the preprocessing module, and combining each second feature information sample to determine multiple second combination information corresponding to each first sample object; and constructing multiple third predictive diagnosis models, and using the multiple second feature information samples corresponding to each first sample object and the corresponding multiple second combination information to train the multiple third predictive diagnosis models respectively.

[0182] Optionally, the predictive diagnosis module also includes: an evaluation module, wherein the evaluation module is connected to the model training module, and the evaluation module is configured to perform the following operations: receiving multiple fourth feature information samples corresponding to each second sample object sent by the preprocessing module, and combining each fourth feature information sample to determine multiple third combination information corresponding to each second sample object; and using the multiple fourth feature information samples corresponding to each second sample object and the multiple third combination information corresponding to each second sample object, determine multiple fourth predictive diagnosis models corresponding to each fourth feature information sample and each third combination information, respectively, wherein the fourth predictive diagnosis model is used to indicate the optimal model corresponding to the fourth feature information sample and the third combination information.

[0183] Therefore, according to this embodiment, the technical effect of improving the accuracy of diagnosis and providing doctors with more accurate and comprehensive decision support can be achieved.

[0184] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0185] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0186] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0187] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0188] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0189] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0190] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A differential diagnosis system for ileocecal ulcer, characterized in that: include: Database, prediction and diagnosis module and output module, The database is connected to the predictive diagnosis module and is configured to provide first disease diagnosis information corresponding to the target object; The output module is connected to the predictive diagnosis module and is configured to output predictive diagnosis information of the target object, wherein the predictive diagnosis information is used to indicate that the target object suffers from Crohn's disease, intestinal tuberculosis or intestinal Behcet's disease, and wherein The predictive diagnostic module is configured to perform the following operations: extracting a plurality of first feature information based on the received first disease diagnosis information corresponding to the target object; Preprocessing the plurality of first feature information to generate a plurality of second feature information, and combining the respective second feature information to determine one or more first combination information; Determine a first prediction and diagnosis model corresponding to each of the second feature information, and use the first prediction and diagnosis model and the corresponding second feature information to perform prediction, thereby generating a first prediction result; Determine a second prediction and diagnosis model corresponding to the one or more first combination information, and use the second prediction and diagnosis model and the corresponding combination information to perform prediction, thereby generating a second prediction result; as well as The first prediction result and the second prediction result are merged to generate the prediction diagnosis information.

2. The system according to claim 1, characterized in that The predictive diagnosis module is also configured to perform the following operations: When a second feature information is extracted, determining a corresponding first prediction diagnosis model; Using the first predictive diagnosis model and the corresponding second feature information to perform prediction, thereby generating a third prediction result; and The third prediction result is used as the predicted diagnosis information.

3. The system according to claim 1, characterized in that The predictive diagnosis module includes: a preprocessing module, wherein the preprocessing module is configured to perform the following operations: The plurality of first feature information are normalized and encoded, and the corresponding plurality of second feature information are generated.

4. The system according to claim 3, characterized in that The operation of fusing the first prediction result and the second prediction result to generate the prediction diagnosis information includes: Determining error rates of respective third predictive diagnostic models in predicting the Crohn's disease, the intestinal tuberculosis, and the intestinal Behcet's disease, wherein the respective third predictive diagnostic models include the first predictive diagnostic model and the second predictive diagnostic model; Based on the error rate, determining a total error rate of multiple model combinations corresponding to the first predictive diagnosis model and the second predictive diagnosis model, and based on the total error rate, determining an average error rate corresponding to the different model combinations; Selecting a minimum average error rate, and using the model combination corresponding to the minimum average error rate as the target model combination; and The first prediction result and the second prediction result corresponding to the target model combination are fused to generate the predictive diagnosis information.

5. The system according to claim 4, characterized in that The pre-processing module is also configured to perform the following operations: Acquire a plurality of first feature information samples corresponding to each first sample object from the database, and perform normalization and encoding processing on the plurality of first feature information samples, thereby generating the plurality of second feature information samples corresponding to each first sample object; as well as A plurality of third feature information samples corresponding to each second sample object are acquired from the database, and the plurality of third feature information samples are normalized and encoded, thereby generating the plurality of fourth feature information samples corresponding to each second sample object.

6. The system according to claim 5, characterized in that The prediction and diagnosis module further includes: a model training module, wherein the model training module is in communication with the preprocessing module and is configured to perform the following operations: receiving a plurality of second feature information samples corresponding to the respective first sample objects sent by the preprocessing module, and combining the respective second feature information samples to determine a plurality of second combination information corresponding to the respective first sample objects; and A plurality of third prediction and diagnosis models are constructed, and the plurality of second characteristic information samples corresponding to the respective first sample objects and the corresponding plurality of second combination information are used to train the plurality of third prediction and diagnosis models respectively.

7. The system according to claim 6, characterized in that The prediction and diagnosis module further includes: an evaluation module, wherein the evaluation module is connected to the model training module, and the evaluation module is configured to perform the following operations: receiving a plurality of fourth feature information samples corresponding to the respective second sample objects sent by the preprocessing module, and combining the respective fourth feature information samples to determine a plurality of third combination information corresponding to the respective second sample objects; and Utilizing multiple fourth feature information samples corresponding to the respective second sample objects and multiple third combination information corresponding to the respective second sample objects, multiple fourth predictive diagnostic models corresponding to the respective fourth feature information samples and the respective third combination information are determined, wherein the fourth predictive diagnostic model is used to indicate the optimal model corresponding to the fourth feature information sample and the third combination information.

8. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is executed, the processor performs the following steps: extracting a plurality of first feature information based on the received first disease diagnosis information corresponding to the target object; Preprocessing the plurality of first feature information to generate a plurality of second feature information, and combining the respective second feature information to determine one or more first combination information; Determine a first prediction and diagnosis model corresponding to each of the second feature information, and use the first prediction and diagnosis model and the corresponding second feature information to perform prediction, thereby generating a first prediction result; Determining a second predictive diagnosis model corresponding to the one or more first combination information, and performing prediction using the second predictive diagnosis model and the corresponding combination information, thereby generating a second prediction result; and The first prediction result and the second prediction result are merged to generate the prediction diagnosis information.

9. A differential diagnosis device for ileocecal ulcer, characterized in that: include: A feature information extraction module, configured to extract a plurality of first feature information based on the received first disease diagnosis information corresponding to the target object; A feature information combination module, used to pre-process the plurality of first feature information, generate a plurality of second feature information, and combine the respective second feature information to determine one or more first combination information; A first prediction result generating module, used for determining first prediction diagnosis models corresponding to the respective second feature information, and performing prediction using the first prediction diagnosis models and the corresponding second feature information, thereby generating a first prediction result; a second prediction result generating module, configured to determine a second prediction diagnosis model corresponding to the one or more first combination information, and to perform prediction using the second prediction diagnosis model and the corresponding combination information, thereby generating a second prediction result; and A prediction result fusion module is used to fuse the first prediction result and the second prediction result to generate the prediction diagnosis information.

10. A differential diagnosis device for ileocecal ulcer, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: extracting a plurality of first feature information based on the received first disease diagnosis information corresponding to the target object; Preprocessing the plurality of first feature information to generate a plurality of second feature information, and combining the respective second feature information to determine one or more first combination information; Determine a first prediction and diagnosis model corresponding to each of the second feature information, and use the first prediction and diagnosis model and the corresponding second feature information to perform prediction, thereby generating a first prediction result; Determine a second prediction and diagnosis model corresponding to the one or more first combination information, and use the second prediction and diagnosis model and the corresponding combination information to perform prediction, thereby generating a second prediction result; as well as The first prediction result and the second prediction result are merged to generate the prediction diagnosis information.

Images