Thyroid nodule clinical feature filling and benign and malignant prediction optimization method and device

Abstract

The invention provides a thyroid nodule clinical feature filling and benign and malignant prediction optimization method and device. The thyroid nodule clinical feature filling and benign and malignant prediction optimization method comprises the following steps of performing binning on each thyroid clinical feature based on a decision tree to obtain an initial binning result; performing binning again on each thyroid nodule clinical feature to obtain an optimal binning result; taking the morbidity rate of the malignant thyroid nodules in each sub-box as sub-box meta-features, obtaining sub-box meta-features of clinical features of all types of thyroid nodules of multiple patients as a training sample set, and training a preset nodule prediction framework by using the training sample set to obtain a nodule prediction model; and using the nodule prediction model to predict the malignant thyroid nodule disease probability of a patient to be detected. According to the scheme, optimal binning is carried out by using importance parameters of clinical features of different thyroid nodules, so that the data complexity is reduced, key information is highlighted, and the malignant thyroid nodules can be predicted more accurately.

Description

Technical Field

[0001] The present application relates to the field of machine learning, and in particular to a method and device for optimizing the filling of clinical features and benign and malignant prediction of thyroid nodules. Background Art

[0002] As a common clinical condition, differentiating between benign and malignant thyroid nodules is crucial for diagnosis and treatment decisions. With the development of medical informatization, clinicians have accumulated a large amount of multi-dimensional diagnostic and treatment data, including ultrasound images, blood biochemical indicators, and patient demographic information. However, this type of data is generally missing features, with a missing rate of 15%-50%, mainly due to the selective implementation of examination items, equipment differences, and omissions in electronic medical record entry. The systematic absence of key predictive features such as ultrasound calcification characteristics and elastography scores has severely restricted the development of medical-assisted diagnosis models based on machine learning. Traditional processing methods can easily lead to significant degradation of model prediction performance.

[0003] Current mainstream methods for addressing missing features in medical clinical data processing have significant limitations. Traditional statistical imputation methods, such as mean / median imputation, ignore nonlinear relationships between features. For example, the threshold effect between thyroglobulin levels and nodule malignancy probability can distort feature distributions. While multiple imputation improves robustness by constructing multiple imputation models, the Markov Chain Monte Carlo method it relies on exponentially increases computational complexity for high-dimensional features, leading to significant performance bottlenecks in thyroid datasets containing 20 or more features. Directly deleting samples with missing values ​​can easily lead to the loss of critical case data in the context of thyroid malignant nodule detection, where positive samples are scarce. Experiments have shown that when the feature missing rate exceeds 20%, this method reduces the training set size by over 40%, significantly reducing model sensitivity. Fixed threshold binning with equal width / frequency binning is difficult to adapt to the multimodal distribution of thyroid features. For example, the interaction between ultrasound TI-RADS grade and nodule size is highly region-specific, and fixed binning can destroy the synergistic predictive signal between features. In addition, existing studies have mostly focused on verifying the imputation effect of a single prediction model, ignoring the differences in feature sensitivity among different models. For example, the tolerance of neural networks to continuous feature imputation errors is significantly lower than that of tree models. Existing methods lack a systematic evaluation framework for multi-model robustness, making them difficult to adapt to clinical multi-model application scenarios. Summary of the Invention

[0004] The embodiment of the present application provides a method and device for filling in the clinical characteristics of thyroid nodules and optimizing the prediction of benign and malignant types. The method uses the importance parameters of different clinical characteristics of thyroid nodules to perform optimal binning, and uses the prevalence of malignant thyroid nodules in each bin as the bin meta-feature to train the preset nodule prediction architecture, thereby reducing data complexity, highlighting key information, addressing the problem of missing values, and enhancing the generalization ability of the model, so that malignant thyroid nodules can be predicted more accurately.

[0005] In a first aspect, the present invention provides a method for optimizing clinical feature filling and benign and malignant prediction of thyroid nodules, the method comprising: Based on the decision tree, the initial binning boundary of each thyroid nodule clinical feature is obtained, and each thyroid clinical feature is binned using the corresponding initial binning boundary to obtain the initial binning result; The importance parameter of each thyroid nodule clinical feature is calculated based on the corresponding initial binning results. The optimal binning boundary of each thyroid nodule clinical feature is adjusted based on the corresponding importance parameter and the initial binning results. Each thyroid nodule clinical feature is re-binned using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity. The prevalence of malignant thyroid nodules in each bin is used as a bin meta-feature to obtain bin meta-features of clinical characteristics of all types of thyroid nodules in multiple patients. The missing bin meta-features of the patients are supplemented as a training sample set. The pre-set nodule prediction framework is trained with the training sample set to obtain a nodule prediction model. The binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested are obtained as the input features to be tested, and the input features to be tested are input into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

[0006] In a second aspect, an embodiment of the present application provides a device for optimizing clinical feature filling and benign and malignant prediction of thyroid nodules, comprising: The initial binning module obtains the initial binning boundaries of each thyroid nodule clinical feature based on the decision tree, and uses the corresponding initial binning boundaries to bin each thyroid clinical feature to obtain the initial binning results; The binning optimization module calculates the importance parameter of each thyroid nodule clinical feature based on the corresponding initial binning results, adjusts the optimal binning boundary of each thyroid nodule clinical feature based on the corresponding importance parameter and the initial binning results, and re-bins each thyroid nodule clinical feature using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity; A training module is configured to use the prevalence of malignant thyroid nodules in each bin as a bin meta-feature, obtain bin meta-features of clinical characteristics of all types of thyroid nodules for multiple patients, supplement the missing bin meta-features of the patients as a training sample set, and train a preset nodule prediction architecture with the training sample set to obtain a nodule prediction model; The prediction module is used to obtain the binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested as the input features to be tested, and input the input features to be tested into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

[0007] In a third aspect, an embodiment of the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a method for optimizing clinical feature filling and benign and malignant prediction of thyroid nodules.

[0008] In a fourth aspect, an embodiment of the present application provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute a process. When the program code is executed by a processor, a method for optimizing the clinical feature filling and benign and malignant prediction of thyroid nodules is implemented.

[0009] The main contributions and innovations of the present invention are as follows: The embodiment of the present application automatically adjusts the binning granularity according to the importance of the features, simultaneously completes feature discretization and standardization, retains the diagnostic value of nonlinear features, and improves the accuracy and generalization of the prediction of benign and malignant thyroid nodules without increasing the computational complexity; the embodiment of the present application addresses the problem of cross-center sample imbalance, breaks through the traditional single-scenario single evaluation quantification path, promotes cross-center feature standardization, and can dynamically select standardization technology based on the center sample size and feature missing distribution, construct a model sensitivity matrix to quantify robustness, and combine relevant means to improve the prediction performance of low-quality center data, thereby ensuring the prediction accuracy for small hospitals; the embodiment of the present application integrates the model prediction results with the doctor's clinical experience decision-making, integrates hierarchical interpretable modules and clinical rules into medical clinical modeling, and improves the doctor's trust and AI-assisted decision-making efficiency.

[0010] The details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1This is a flow chart of a method for optimizing the clinical feature filling and benign and malignant prediction of thyroid nodules according to an embodiment of the present application; Figure 2 This is a structural block diagram of a device for optimizing clinical feature filling and benign and malignant prediction of thyroid nodules according to an embodiment of the present application; Figure 3 Schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0012] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The implementations described in the following exemplary embodiments are not intended to represent all implementations consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with certain aspects of one or more embodiments of this specification, as detailed in the appended claims.

[0013] It should be noted that in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the method may include more or fewer steps than those described in this specification. In addition, a single step described in this specification may be broken down into multiple steps for description in other embodiments, and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0014] Example 1 The embodiment of the present application provides a method for filling clinical features of thyroid nodules and optimizing the prediction of benign and malignant thyroid nodules, which uses the importance parameters of different clinical features of thyroid nodules to perform optimal binning, and uses the prevalence of malignant thyroid nodules in each bin as the bin meta-feature to train the preset nodule prediction architecture, thereby reducing data complexity, highlighting key information, dealing with missing value problems and enhancing the generalization ability of the model, and can more accurately predict malignant thyroid nodules. Specifically, with reference to Figure 1 , the method comprising: Based on the decision tree, the initial binning boundary of each thyroid nodule clinical feature is obtained, and each thyroid clinical feature is binned using the corresponding initial binning boundary to obtain the initial binning result; The importance parameter of each thyroid nodule clinical feature is calculated based on the corresponding initial binning results. The optimal binning boundary of each thyroid nodule clinical feature is adjusted based on the corresponding importance parameter and the initial binning results. Each thyroid nodule clinical feature is re-binned using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity. The prevalence of malignant thyroid nodules in each bin is used as a bin meta-feature to obtain bin meta-features of clinical characteristics of all types of thyroid nodules in multiple patients. The missing bin meta-features of the patients are supplemented as a training sample set. The pre-set nodule prediction framework is trained with the training sample set to obtain a nodule prediction model. The binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested are obtained as the input features to be tested, and the input features to be tested are input into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

[0015] In some embodiments, the initial binning boundaries for each thyroid nodule clinical feature are obtained by setting the maximum depth of the decision tree.

[0016] Specifically, before binning, each clinical feature of thyroid nodules is first arranged in order of size, and then each clinical feature of thyroid nodules is binned based on the corresponding initial binning boundary to obtain the initial binning result.

[0017] In some specific embodiments, the clinical characteristics of thyroid nodules include calcium micronization characteristics, aspect ratio characteristics, echo characteristics, boundary characteristics, and blood flow characteristics.

[0018] Specifically, the microcalcification feature refers to the tiny calcification foci that may be produced by cancer cells during the metabolic process. Therefore, under ultrasound imaging, the presence of microcalcification often indicates that the thyroid nodule has a higher risk of malignancy. In contrast, coarse calcification has larger calcification foci and relatively irregular shapes. Coarse calcification may also appear in benign nodules.

[0019] Specifically, the aspect ratio refers to the ratio of the anterior-posterior diameter to the lateral diameter of the nodule in the ultrasound image. When the aspect ratio is greater than 1, the nodule appears to be vertically long, which often indicates that the possibility of malignancy of the nodule is increased; while benign nodules tend to be more horizontally long, with an aspect ratio of less than 1, and the overall shape is relatively flat and round.

[0020] Specifically, echo characteristics are one of the key observation points in ultrasound examination of thyroid nodules. Hypoechoic nodules have a relatively higher risk of malignancy, while hyperechoic nodules are more common in benign lesions.

[0021] Specifically, the clarity of a nodule's boundaries is crucial for determining its nature. Benign nodules generally have clear boundaries that can be clearly distinguished from surrounding thyroid tissue, with smooth edges, indicating relatively regular growth and no irregular infiltration into surrounding tissue. Malignant nodules, on the other hand, often have unclear boundaries, often appearing fuzzy and with unclear demarcation from surrounding tissue.

[0022] Specifically, observing the blood flow signals inside and around the nodule through ultrasound can also help determine the nature of the nodule. The blood flow signals of benign nodules are usually relatively regular and uniform, and are mostly peripheral. Malignant nodules have active cell proliferation and require more nutrients, so they often have richer and more chaotic blood flow signals. Not only does the peripheral blood flow increase, but more branched and penetrating blood flow distribution can also be seen inside.

[0023] In some specific embodiments, the importance score of each bin in the initial binning results is calculated separately, and the importance score of each bin under the same thyroid nodule clinical feature is integrated to obtain the importance parameter of each thyroid nodule clinical feature.

[0024] Specifically, this solution uses IV (information value) to calculate the importance score of each bin. The formula of the importance parameter is as follows:

[0025] in, Clinical features of thyroid nodules The importance parameter, is the importance score of bin i, and k is the clinical characteristics of thyroid nodules The number of bins.

[0026] Furthermore, the information gain of the clinical characteristics of thyroid nodules under the current binning boundary is calculated, and the information gain and the importance parameter are weighted and summed to obtain the information gain feature corresponding to the clinical characteristics of thyroid nodules. The current binning boundary when the information gain feature is maximum is used as the optimal binning boundary corresponding to the clinical characteristics of thyroid nodules.

[0027] It is worth mentioning that the maximum value of the information gain feature is found by traversal iteration. In the first round of traversal iteration, the current binning boundary is the initial binning boundary. In subsequent iterations, the current binning boundary is continuously updated until the optimal binning boundary is obtained.

[0028] Specifically, the formula for obtaining the optimal bin boundary is as follows:

[0029] in, and is the balance coefficient, which is used to balance the influence of information gain and importance parameters. Clinical features of thyroid nodules The information gain function, Clinical features of thyroid nodules The importance parameter of B is the current bin boundary, B={ }.

[0030] Specifically, this scheme comprehensively considers information gain and importance parameters, and uses the balance coefficient to balance the influence of the two, avoiding the one-sidedness of selecting bin boundaries based solely on information gain or feature importance. It can more scientifically and comprehensively measure the value of clinical characteristics of thyroid nodules in different binning situations, making the selected bin boundaries more in line with actual application needs.

[0031] Furthermore, in this scheme, the importance parameter of the clinical characteristics of thyroid nodules is positively correlated with the granularity of binning. That is, the more important the clinical characteristics of thyroid nodules are for judging benign or malignant, the finer the granularity of binning.

[0032] In some specific embodiments, the number of patients with malignant thyroid nodules in each bin is obtained, and the ratio of the number of patients with malignant thyroid nodules in the bin to the number of all patients in the bin is used as the bin meta-feature.

[0033] Specifically, the formula for binning meta-features is as follows:

[0034] in, is the bin meta-feature, m is the number of patients with malignant thyroid nodules in the bin, and N is the total number of patients in the bin.

[0035] Specifically, the data used in the training of this program is retrieved from various hospital centers, so the benign or malignant status of each patient's thyroid nodule is diagnosed and marked by the doctor.

[0036] In some specific embodiments, due to the diverse and complex nature of thyroid clinical characteristics, using the original data directly as model input can lead to significant deviations in prediction results and easily lead to overfitting. This approach, however, uses the prevalence of malignant thyroid nodules within each bin as the bin meta-feature, summarizing and simplifying the original complex clinical characteristic data. By integrating the previously scattered and complex data into a relatively unified metric of prevalence corresponding to different bins, the input data becomes more organized and concise, helping to reduce data complexity and enabling subsequent model training to more efficiently focus on key information, avoiding problems such as overfitting due to overly complex data.

[0037] In addition, this scheme focuses the binning meta-features on the prevalence of malignant thyroid nodules in each bin. This indicator is directly related to the key goal of judging whether thyroid nodules are benign or malignant. In this way, it can highlight the relationship between different clinical characteristics and whether the nodules are malignant in different bins, making it easier for the model to capture those feature patterns that are important for judging the nature of nodules during training, thereby enhancing the model's ability to distinguish benign and malignant nodules, and more accurately digging out effective predictive information hidden in many clinical characteristics.

[0038] In some specific embodiments, missing patients who are missing the same type of bin meta-features are placed in a set as a missing patient set, and the ratio of the number of patients with malignant thyroid nodules in the missing patient set to the number of all missing patients in the bin is used as a supplementary bin meta-feature, and the supplementary bin meta-feature is used to complete the corresponding type of bin meta-features of each missing patient in the corresponding missing patient set.

[0039] For example, for the case of missing age features, all patients with missing ages are put into a set as the age-missing patient set, and then the prevalence of the number of patients with malignant thyroid nodules in the age-missing patient set is used as a supplementary bin meta-feature. If it is 0.6, then the age bin meta-feature in the age-missing patient set is supplemented to 0.6.

[0040] In some specific embodiments, during the training of the nodule prediction architecture, the weights of the misjudged samples are dynamically adjusted, and the weights of the misjudged samples are introduced into the prediction loss function of the nodule prediction architecture. The formula for adjusting the weights of the misjudged samples is expressed as follows:

[0041] in, is the weight of the i-th sample in the training sample set at the t+1th iteration, is the weight of the i-th sample in the training sample set at the t-th iteration, is an indicator function, which is 1 when a misjudgment occurs and 0 otherwise. is the weight adjustment coefficient, , is the weighted error rate of the t-th round model, To compensate for the hyperparameters.

[0042] The formula for introducing the weight of misclassified samples into the prediction loss function of the nodule prediction architecture is as follows:

[0043] in, is the prediction loss function, is the weight of the i-th sample in the training sample set at the t-th iteration, is the prediction result of the i-th sample.

[0044] Specifically, by dynamically adjusting the weights of misjudged samples and introducing the weights of misjudged samples into the prediction loss function of the nodule prediction architecture, the model can be guided to focus on learning misjudged samples, prompting the model to adjust its own parameters in the direction of more accurately handling such misjudgment situations, ultimately achieving the effect of improving model accuracy and generalization ability.

[0045] In some specific embodiments, multiple types of nodule prediction architectures are constructed, and multiple types of nodule prediction architectures are trained simultaneously using training data sets to obtain multiple types of nodule prediction models. A prediction sensitivity matrix is ​​constructed for each nodule prediction model, and the prediction sensitivity matrix is ​​used to evaluate the prediction accuracy of the corresponding nodule prediction model under different feature missing rates, wherein the feature missing rate represents the missing status of clinical features of thyroid nodules. Based on the feature missing rates in different hospital center databases, the nodule prediction model with the highest prediction accuracy is selected to predict the probability of malignant thyroid nodules in the patient to be tested.

[0046] Specifically, due to different medical practices in different hospitals, there will be certain differences in the items of examination for thyroid nodules. Some hospitals will check all clinical characteristics of thyroid nodules including calcium micro-formation characteristics, aspect ratio characteristics, echo characteristics, boundary characteristics and blood flow characteristics, while some hospitals may only check calcium micro-formation characteristics, aspect ratio characteristics, echo characteristics and boundary characteristics. For nodule prediction models, when there are uncertain feature missings, different types of models may be affected differently. For example, the XGboost model has an accuracy of 0.95 for feature prediction with a missing rate of 0, but an accuracy of 0.78 for feature prediction with a missing rate of 30%. In the logistic regression model, the accuracy of feature prediction with a missing rate of 0 is 0.91, and the accuracy of feature prediction with a missing rate of 30% is 0.85.

[0047] Furthermore, the sensitivity matrix is ​​expressed as , the elements in the sensitivity matrix are , Represents the relative performance value of the indicator vector q of the nodule prediction model m under the missing rate k.

[0048] Specifically, represents the indicator vector of the nodule prediction model m on the dataset with a missing rate of 0, Represents the indicator vector of nodule prediction model m on the dataset with missing rate k.

[0049] Specifically, The calculation formula is:

[0050] in, For the nodule prediction model m in the dataset with missing rate k Indicators for prediction, where: For the nodule prediction model m in the dataset with missing rate k The sensitivity of the prediction is to evaluate the proportion of true positive (malignant nodules) samples in the data set that are predicted to be positive by the model. The calculation formula is:

[0051] Here, represents the number of samples whose true label is positive and predicted as positive, and represents the number of samples whose true label is positive but predicted as negative. Sensitivity represents the recall rate of positive samples. If a positive sample is not correctly predicted, it is considered a "missed diagnosis."

[0052] For the nodule prediction model m in the dataset with missing rate k The specificity of the prediction is the proportion of true negative (benign nodules) samples predicted by the model as negative, and the calculation formula is:

[0053] Here, represents the number of samples predicted as negative when the true label is negative, and represents the number of samples predicted as positive when the true label is negative. Specificity represents the recall rate of negative samples. If a negative sample is not correctly predicted, it is considered a "misdiagnosis."

[0054] For the nodule prediction model m in the dataset with missing rate k The area enclosed by the ROC curve and the false positive rate abscissa when making predictions evaluates the model's ability to distinguish between positive (positive / malignant) and negative (negative / benign) classes. The larger the AUC, the stronger the model's ability to distinguish. The AUC calculation formula is:

[0055]

[0056]

[0057]

[0058] Among them, represents the number of patient samples with positive thyroid nodules in the dataset, represents the number of patient samples with negative thyroid nodules in the dataset, and represents the output score of the prediction model, which represents the indicator function of a yin-yang sample pair composed of a group of negative samples and positive samples.

[0059] For the nodule prediction model m in the dataset with missing rate k The accuracy of prediction is used to evaluate the proportion of all correctly predicted samples to the total number of samples. The calculation formula is:

[0060] For the nodule prediction model m in the dataset with missing rate k The accuracy rate when making predictions on ) and the harmonic mean of recall rate (Recall), which is a comprehensive evaluation indicator around positive samples. Its calculation formula is:

[0061]

[0062]

[0063] For the nodule prediction model m in the dataset with missing rate k The positive predictive value of the prediction is equivalent to the precision, which evaluates the proportion of samples predicted to be positive whose true labels are positive. The calculation formula is:

[0064] For the nodule prediction model m in the dataset with missing rate k The negative predictive value of the prediction is to evaluate the proportion of samples predicted to be negative whose true label is negative. The calculation formula is:

[0065] In some specific embodiments, an interpretable method is used to quantify the influence of each binned meta-feature of the patient to be tested on the predicted probability of malignant thyroid nodules, and a waterfall chart is generated to intuitively show how each feature affects the probability of malignancy (such as microcalcification contribution +23%), and a decision tree is combined to visualize the key judgment path (such as "aspect ratio > 1 → malignancy probability increases by 18%").

[0066] Furthermore, the generated waterfall chart is combined with the thyroid clinical guidelines to generate a structured report template, and clinicians correct or confirm the probability of malignant thyroid nodules based on the structured report template.

[0067] Example 2 Based on the same concept, Figure 2This application also proposes a device for optimizing the clinical characteristics of thyroid nodules and their benign and malignant prediction, including: The initial binning module obtains the initial binning boundaries of each thyroid nodule clinical feature based on the decision tree, and uses the corresponding initial binning boundaries to bin each thyroid clinical feature to obtain the initial binning results; The binning optimization module calculates the importance parameter of each thyroid nodule clinical feature based on the corresponding initial binning results, adjusts the optimal binning boundary of each thyroid nodule clinical feature based on the corresponding importance parameter and the initial binning results, and re-bins each thyroid nodule clinical feature using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity; A training module is configured to use the prevalence of malignant thyroid nodules in each bin as a bin meta-feature, obtain bin meta-features of clinical characteristics of all types of thyroid nodules for multiple patients, supplement the missing bin meta-features of the patients as a training sample set, and train a preset nodule prediction architecture with the training sample set to obtain a nodule prediction model; The prediction module is used to obtain the binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested as the input features to be tested, and input the input features to be tested into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

[0068] Example 3 This embodiment also provides an electronic device, referring to Figure 3 , includes a memory 404 and a processor 402, wherein the memory 404 stores a computer program, and the processor 402 is configured to run the computer program to perform the steps in any of the above method embodiments.

[0069] Specifically, the processor 402 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0070] Memory 404 may include a large-capacity memory 404 for data or instructions. By way of example, and not limitation, memory 404 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 404 may include removable or non-removable (or fixed) media. Where appropriate, memory 404 may be internal or external to the data processing device. In certain embodiments, memory 404 is non-volatile memory. In certain embodiments, memory 404 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM) or a flash memory (FLASH), or a combination of two or more of these. In appropriate circumstances, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM may be a fast page mode dynamic random access memory 404 (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0071] The memory 404 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 402 .

[0072] The processor 402 reads and executes computer program instructions stored in the memory 404 to implement any one of the methods for optimizing the clinical feature filling and benign and malignant prediction of thyroid nodules in the above embodiments.

[0073] Optionally, the electronic device may further include a transmission device 406 and an input / output device 408 , wherein the transmission device 406 is connected to the processor 402 , and the input / output device 408 is connected to the processor 402 .

[0074] Transmission device 406 can be used to receive or transmit data via a network. Specific examples of such networks may include wired or wireless networks provided by the electronic device's communications provider. In one embodiment, the transmission device includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 406 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0075] The input / output device 408 is used to input or output information. In this embodiment, the input information may be various clinical characteristics of thyroid nodules, and the output information may be the probability of malignant thyroid nodules.

[0076] Optionally, in this embodiment, the processor 402 may be configured to execute the following steps through a computer program: Based on the decision tree, the initial binning boundary of each thyroid nodule clinical feature is obtained, and each thyroid clinical feature is binned using the corresponding initial binning boundary to obtain the initial binning result; The importance parameter of each thyroid nodule clinical feature is calculated based on the corresponding initial binning results. The optimal binning boundary of each thyroid nodule clinical feature is adjusted based on the corresponding importance parameter and the initial binning results. Each thyroid nodule clinical feature is re-binned using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity. The prevalence of malignant thyroid nodules in each bin is used as a bin meta-feature to obtain bin meta-features of clinical characteristics of all types of thyroid nodules in multiple patients. The missing bin meta-features of the patients are supplemented as a training sample set. The pre-set nodule prediction framework is trained with the training sample set to obtain a nodule prediction model. The binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested are obtained as the input features to be tested, and the input features to be tested are input into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

[0077] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be repeated here.

[0078] In general, various embodiments may be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flow charts, or using some other graphical representation, it should be understood that, as non-limiting examples, the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or a controller or other computing device, or some combination thereof.

[0079] The embodiments of the present invention may be implemented by computer software that is executable by a data processor of a mobile device, such as in a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros may be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. A computer program product may include one or more computer executable components that are configured to perform an embodiment when the program is run. One or more computer executable components may be at least one software code or a portion thereof. In addition, it should be noted at this point that, for example, Figure 3 Any block of the logic flow in the program may represent program steps, or interconnected logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions. The software may be stored on physical media such as memory chips or memory blocks implemented within the processor, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and their data variants, CDs, etc. Physical media are non-transitory media.

[0080] Those skilled in the art should understand that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0081] The above embodiments merely illustrate several embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules, characterized by: The following steps are involved: Based on the decision tree, the initial binning boundary of each thyroid nodule clinical feature is obtained, and each thyroid clinical feature is binned using the corresponding initial binning boundary to obtain the initial binning result; The importance parameter of each thyroid nodule clinical feature is calculated based on the corresponding initial binning results. The optimal binning boundary of each thyroid nodule clinical feature is adjusted based on the corresponding importance parameter and the initial binning results. Each thyroid nodule clinical feature is re-binned using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity. The prevalence of malignant thyroid nodules in each bin is used as a bin meta-feature to obtain bin meta-features of clinical characteristics of all types of thyroid nodules in multiple patients. The missing bin meta-features of the patients are supplemented as a training sample set. The pre-set nodule prediction framework is trained with the training sample set to obtain a nodule prediction model. The binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested are obtained as the input features to be tested, and the input features to be tested are input into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

2. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules according to claim 1, characterized in that: The importance score of each bin in the initial binning results was calculated separately, and the importance score of each bin under the same clinical feature of thyroid nodules was integrated to obtain the importance parameter of each clinical feature of thyroid nodules.

3. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules according to claim 1, characterized in that: Calculate the information gain of the clinical characteristics of thyroid nodules under the current binning boundary, perform weighted summation of the information gain and the importance parameter to obtain the information gain feature corresponding to the clinical characteristics of thyroid nodules, and use the current binning boundary when the information gain feature is maximum as the optimal binning boundary corresponding to the clinical characteristics of thyroid nodules.

4. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules according to claim 1, characterized in that: The number of patients with malignant thyroid nodules in each bin was obtained, and the ratio of the number of patients with malignant thyroid nodules in the bin to the number of all patients in the bin was used as the bin meta-feature.

5. The method for optimizing clinical feature filling and benign and malignant prediction of thyroid nodules according to claim 1, characterized in that: Patients with missing bin meta-features are regarded as missing patients, and missing patients who lack the same type of bin meta-features are placed in a set as a missing patient set. The ratio of the number of patients with malignant thyroid nodules in the missing patient set to the number of all missing patients in the bin is used as the supplementary bin meta-feature, and the supplementary bin meta-feature is used to complete the corresponding type of bin meta-feature of each missing patient in the corresponding missing patient set.

6. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules according to claim 1, characterized in that: During the training of the nodule prediction architecture, the weights of misclassified samples are dynamically adjusted, and the weights of misclassified samples are introduced into the prediction loss function of the nodule prediction architecture. The formula for adjusting the weights of misclassified samples is expressed as follows: ， in, is the weight of the i-th sample in the training sample set at the t+1th iteration, is the weight of the i-th sample in the training sample set at the t-th iteration, is an indicator function, which is 1 when a misjudgment occurs and 0 otherwise. is the weight adjustment coefficient, , is the weighted error rate of the t-th round model, To compensate for hyperparameters; The formula for introducing the weight of misclassified samples into the prediction loss function of the nodule prediction architecture is as follows: ； in, is the prediction loss function, is the weight of the i-th sample in the training sample set at the t-th iteration, is the prediction result of the i-th sample.

7. A method for clinical feature filling and benign and malignant prediction optimization of thyroid nodules according to claim 1, characterized in that: Multiple types of nodule prediction architectures are constructed, and multiple types of nodule prediction architectures are trained simultaneously using training data sets to obtain multiple types of nodule prediction models. A prediction sensitivity matrix is ​​constructed for each nodule prediction model. The prediction sensitivity matrix is ​​used to evaluate the prediction accuracy of the corresponding nodule prediction model under different feature missing rates, wherein the feature missing rate represents the missing status of clinical features of thyroid nodules. Based on the feature missing rates in different hospital center databases, the nodule prediction model with the highest prediction accuracy is selected to predict the probability of malignant thyroid nodules in the patient to be tested.

8. A device for clinical feature filling and benign and malignant prediction optimization of thyroid nodules, characterized by: include: The initial binning module obtains the initial binning boundaries of each thyroid nodule clinical feature based on the decision tree, and uses the corresponding initial binning boundaries to bin each thyroid clinical feature to obtain the initial binning results; The binning optimization module calculates the importance parameter of each thyroid nodule clinical feature based on the corresponding initial binning results, adjusts the optimal binning boundary of each thyroid nodule clinical feature based on the corresponding importance parameter and the initial binning results, and re-bins each thyroid nodule clinical feature using the corresponding optimal binning boundary to obtain the optimal binning result. The importance parameter is positively correlated with the binning granularity; A training module is configured to use the prevalence of malignant thyroid nodules in each bin as a bin meta-feature, obtain bin meta-features of clinical characteristics of all types of thyroid nodules for multiple patients, supplement the missing bin meta-features of the patients as a training sample set, and train a preset nodule prediction architecture with the training sample set to obtain a nodule prediction model; The prediction module is used to obtain the binned meta-features of multiple thyroid nodule clinical characteristics of the patient to be tested as the input features to be tested, and input the input features to be tested into the nodule prediction model to obtain the probability of malignant thyroid nodules in the patient to be tested.

9. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the method for optimizing the clinical feature filling and benign and malignant prediction of thyroid nodules according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which includes program code for controlling a process to execute a process. When the program code is executed by a processor, a method for filling in clinical characteristics of thyroid nodules and optimizing benign and malignant prediction is implemented.

Images