Thyroid papillary carcinoma recurrence prediction method and device applied to postoperative patient

Abstract

The embodiment of the invention discloses a papillary thyroid carcinoma recurrence prediction method and device applied to postoperative patients. A specific embodiment of the method comprises the following steps: generating emotional state information and sleep state information for a postoperative patient according to physiological data corresponding to the postoperative patient; in response to emotional state information representing emotional abnormity and / or sleep state information representing sleep state abnormity, acquiring self-evaluation data corresponding to the postoperative patient through a dynamically generated data acquisition scale; generating a recurrence prediction result according to the multi-modal inspection data associated with the postoperative patient, the self-evaluation data and a pre-trained papillary thyroid carcinoma recurrence prediction model; and according to a trigger condition triggered by the recurrence prediction result, initiating a recurrence risk prompt to a trigger condition associated with the trigger condition. By means of the implementation mode, recurrence prediction of the thyroid papillary carcinoma postoperative patient is effectively achieved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the fields of computer technology and medical health technology, and particularly to a method and device for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients. Background Art

[0002] Thyroid cancer is a malignant tumor originating from thyroid follicular epithelial or parafollicular epithelial cells, and is one of the common malignant tumors in the head and neck. Among them, papillary thyroid carcinoma (PTC) is a common subtype of thyroid cancer, and its incidence accounts for about 75.5% - 87.3% of thyroid cancer. After standardized clinical comprehensive treatment such as surgical operation, microwave ablation, and hormone therapy for PTC patients, the 10-year survival rate is 80% - 90%; however, 5% - 30% of the patients are still prone to recurrence or cervical lymph node metastasis, invade surrounding tissues, and affect the quality of life. Therefore, predicting the recurrence of PTC after surgery and intervening in a timely manner has important clinical significance.

[0003] Currently, the intervention means for clinical prevention of recurrence mainly rely on traditional medical models such as surgical thoroughness assessment, radioactive iodine therapy, and TSH suppression therapy. These methods have limitations such as large individual response differences and poor tolerance to secondary treatment. It is relatively difficult to determine the impact of postoperative thyroid hormone and antibody levels on the prognosis of PTC. Especially for patients after lobectomy and thermal ablation, although it is reasonable that the recurrence risk increases with the increase of TSH theoretically, due to the influence of the function of the residual thyroid, the relationship between TSH and postoperative recurrence is complex. Although postoperative patients can receive TSH suppression therapy regularly, the thyroid-related hormone levels and antibody levels are still affected by various complex factors, which may increase the recurrence risk of thyroid cancer.

[0004] The above information disclosed in this background art section is only used to enhance the understanding of the background of the inventive concept, and thus, it may include information that does not form the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] This summary of the present disclosure is used to introduce concepts in a brief form, and these concepts will be described in detail in the following detailed implementation section. This summary of the present disclosure is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to be used to limit the scope of the claimed technical solution.

[0006] Some embodiments of the present disclosure propose a method and device for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients to solve the technical problems mentioned in the above background art section.

[0007] In a first aspect, some embodiments of the present disclosure provide a method for predicting recurrence of papillary thyroid carcinoma for postoperative patients, the method comprising: generating, based on the physiological data corresponding to a postoperative patient, emotional state information and sleep state information for the postoperative patient, wherein the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time; in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collecting, through a dynamically generated data collection scale, self-evaluation data corresponding to the postoperative patient, wherein the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; generating a recurrence prediction result based on multimodal examination test data associated with the postoperative patient, the self-evaluation data, and a pre-trained papillary thyroid carcinoma recurrence prediction model, the recurrence prediction result including: a recurrence risk type and a recurrence risk confidence level; and initiating a recurrence risk prompt associated with the trigger condition according to the trigger condition triggered by the recurrence prediction result.

[0008] In a second aspect, some embodiments of the present disclosure provide a device for predicting recurrence of papillary thyroid carcinoma for postoperative patients, the device comprising: a first generation unit configured to generate, based on the physiological data corresponding to a postoperative patient, emotional state information and sleep state information for the postoperative patient, wherein the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time; a collection unit configured to, in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collect, through a dynamically generated data collection scale, self-evaluation data corresponding to the postoperative patient, wherein the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; a second generation unit configured to generate a recurrence prediction result based on multimodal examination test data associated with the postoperative patient, the self-evaluation data, and a pre-trained papillary thyroid carcinoma recurrence prediction model, the recurrence prediction result including: a recurrence risk type and a recurrence risk confidence level; and an initiation unit configured to initiate a recurrence risk prompt associated with the trigger condition according to the trigger condition triggered by the recurrence prediction result.

[0009] In a third aspect, some embodiments of the present disclosure provide an electronic device, comprising: one or more processors; a storage device having stored thereon one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the method described in any implementation manner of the first aspect above.

[0010] Fourthly, some embodiments of the present disclosure provide a computer-readable medium, on which a computer program is stored. When the program is executed by a processor, the method described in any implementation manner of the above first aspect is implemented.

[0011] The above various embodiments of the present disclosure have the following beneficial effects: Through the method for predicting the recurrence of papillary thyroid cancer applied to postoperative patients in some embodiments of the present disclosure, the recurrence prediction for postoperative patients with papillary thyroid cancer is effectively realized. Specifically, first, according to the physiological data corresponding to the postoperative patients, the emotional state information and sleep state information for the above postoperative patients are generated, where the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time. In practice, there is still a certain recurrence risk of papillary thyroid cancer after surgery. At the same time, the emotional state and sleep state of postoperative patients may also have a significant impact on thyroid function, thereby affecting the recurrence of the thyroid. Therefore, the present disclosure generally obtains the physiological data of postoperative patients, and then judges the emotional state and sleep state of postoperative patients. Secondly, in response to the fact that the above emotional state information characterizes abnormal emotions and / or the above sleep state information characterizes abnormal sleep states, through a dynamically generated data collection scale, the self-evaluation data corresponding to the above postoperative patients is collected, where the above self-evaluation data characterizes the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression. In practice, since the recurrence time varies among individuals, the method of regularly collecting data using a scale may delay the discovery of recurrence. At the same time, too high a data collection frequency may also cause the patient's resistance. In addition, conventional scales often use a fixed format. Especially at a certain data collection frequency, patients are prone to the problem of habitual filling, resulting in inaccurate self-evaluation data collected. Therefore, the present disclosure determines the timing of data collection by combining the emotional state and sleep state to determine the combined scale, and dynamically generates the scale, so as to dynamically adjust the collection time and collection method according to the individual situation. Then, according to the multimodal examination and test data associated with the above postoperative patients, the above self-evaluation data, and a pre-trained recurrence prediction model for papillary thyroid cancer, a recurrence prediction result is generated, and the above recurrence prediction result includes: recurrence risk type and recurrence risk confidence level. By combining self-evaluation data, examination and test data, and a prediction model, accurate prediction of recurrence is thus achieved. Finally, according to the trigger condition triggered by the above recurrence prediction result, a recurrence risk prompt is sent to the initiator associated with the above trigger condition. In this way, the risk prompt is automatically triggered, so that patients can seek medical treatment in time when they have a recurrence risk. Through this method, the recurrence prediction for postoperative patients with papillary thyroid cancer is effectively realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In combination with the accompanying drawings and with reference to the following specific embodiments, the above and other features, advantages and aspects of the various embodiments of the present disclosure will become more apparent. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic and the elements and elements are not necessarily drawn to scale.

[0013] Figure 1 is a flowchart of some embodiments of a method for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients according to the present disclosure; Figure 2 is a schematic diagram of the relationship between the data acquisition window and the operation time; Figure 3 is a schematic diagram of the generation process of the oxygen saturation signal after filling; Figure 4 is a schematic diagram of the generation process of the emotional state information; Figure 5 is a schematic structural diagram of some embodiments of a device for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients according to the present disclosure; Figure 6 is a schematic structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. Specific Embodiments

[0014] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0015] In addition, it should be noted that for the sake of convenience of description, only the parts related to the relevant invention are shown in the drawings. Without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other.

[0016] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependent relationships.

[0017] It should be noted that the modifications of "one" and "plural" mentioned in the present disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly specified in the context, it should be understood as "one or more".

[0018] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0019] Regarding operations such as the collection, storage, and use of user data (e.g., physiological data, multi-modal examination data, self-evaluation data) involved in the present disclosure, before performing the corresponding operations, relevant organizations or individuals shall fulfill obligations including conducting personal information security impact assessments, fulfilling the obligation of notification to the personal information subject, and obtaining the prior authorization and consent of the personal information subject. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0020] Reference Figure 1 , shows a flow 100 of some embodiments of a method for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients according to the present disclosure. The method for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients includes the following steps: Step 101, generate emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient.

[0021] In some embodiments, the execution subject (e.g., a computing device) of the method for predicting recurrence of papillary thyroid carcinoma applied to postoperative patients can generate emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient. Among them, the physiological data includes: basic physiological data and sleep physiological data. Among them, the basic physiological data represents the physiological index data related to the postoperative patient. The sleep physiological data represents the sleep index data related to the postoperative patient. In practice, the physiological data of the postoperative patient can be continuously collected by a smart wearable device (e.g., a smart watch). The window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time. In addition, steps such as data cleaning and standardization processing can be performed on the physiological data before use, for example, removing duplicate data, outliers, and interpolating missing values.

[0022] As an example, see Figure 2Schematic diagram of the relationship between the data acquisition window and the operation time. Among them, in the time period closer to the operation time, a shorter data acquisition window A is selected. As the time length from the operation time increases, the data acquisition window B is larger than the data acquisition window A, and the data acquisition window C is larger than the data acquisition window B. The reason for using this method for data acquisition is that the conventional method of receiving physiological data sent by intelligent wearable devices in real time will generate a relatively large load pressure on the server side. At the same time, since the recurrence time of papillary thyroid cancer symptoms varies among individuals, the average recurrence cycle of papillary thyroid cancer can be statistically analyzed by combining samples, and the window length of the data acquisition window can be dynamically adjusted accordingly. In addition, as the data acquisition window is farther from the operation time, a certain downsampling method can be used to process the physiological data to avoid the problem of a large amount of data generated due to the increase in the window length of the data acquisition window.

[0023] Optionally, the basic physiological data includes: blood pressure signal, blood oxygen signal, heart rate signal. The sleep physiological data includes: a sleep data sequence, and the sleep data includes: sleep duration, wakefulness duration, light sleep ratio, deep sleep ratio, and sleep distribution. Specifically, the sleep data can be in days as the granularity.

[0024] It should be noted that the above computing device can be hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster composed of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is embodied as software, it can be installed in the above-listed hardware devices. It can be implemented as, for example, multiple software or software modules for providing distributed services, or as a single software or software module. No specific limitation is made here.

[0025] In some optional implementation manners of some embodiments, the above execution subject generates emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient, including: In the first step, abnormal signal localization is respectively performed on the above blood pressure signal, the above blood oxygen signal, and the above heart rate signal to obtain an abnormal blood pressure signal sequence, an abnormal blood oxygen signal sequence, and an abnormal heart rate signal sequence.

[0026] Among them, the abnormal blood pressure signal refers to the local blood pressure signal in the blood pressure signal that is in an abnormal state. The abnormal blood oxygen signal refers to the local blood oxygen signal in the blood oxygen signal that is in an abnormal state. The abnormal heart rate signal refers to the local heart rate signal in the heart rate signal that is in an abnormal state. Specifically, for the blood pressure signal, the abnormal blood pressure signal sequence can be filtered out from the blood pressure signal by setting a low blood pressure threshold and a high blood pressure threshold. For the blood oxygen signal, the abnormal blood oxygen signal sequence can be filtered out from the blood oxygen signal by setting a low blood oxygen threshold. For the heart rate signal, since the heart rate has certain periodic characteristics, the baseline heart rate waveform of the postoperative patient within a single cycle can be determined by statistical means, and the abnormal heart rate signal sequence can be identified by comparing the waveform similarity between the heart rate waveform in each cycle and the baseline heart rate waveform.

[0027] Second, according to the above-mentioned blood pressure signal, the above-mentioned blood oxygen signal, the above-mentioned heart rate signal, the above-mentioned abnormal blood pressure signal sequence, the above-mentioned abnormal blood oxygen signal sequence, and the above-mentioned abnormal heart rate signal sequence, determine the first filling information sequence, the second filling information sequence, and the third filling information sequence.

[0028] Among them, the first filling information includes: the first interpolation and the filling amount. The first interpolation represents the signal interpolation between two adjacent abnormal blood pressure signals. The second filling information includes: the second interpolation and the filling amount. The second interpolation represents the signal interpolation between two adjacent abnormal blood oxygen signals. The third filling information includes: the third interpolation and the filling amount. The third interpolation represents the signal interpolation between two adjacent abnormal heart rate signals.

[0029] As an example, taking the abnormal blood oxygen signal sequence as an example, since the abnormal blood oxygen signal is the local blood oxygen signal in the blood oxygen signal that is in an abnormal state, there is an interval (i.e., normal blood oxygen signal) between different abnormal blood oxygen signals in the time dimension. Since the blood oxygen signal is continuously collected by the smart wearable device, there is a problem of a large amount of data. Conventionally, such as the method of setting the normal blood oxygen signal to 0, although it is relatively simple (the reason is that by setting to 0, the subsequent data processing amount can be reduced by the way of skipping 0), the method of setting to 0 ignores the normal blood oxygen signal characteristics of the postoperative patient. Another example is that the signal length of the blood oxygen signal can be reduced by directly splicing the abnormal blood oxygen signals in the abnormal blood oxygen signal sequence, but the time interval characteristics between any two abnormal blood oxygen signals when the abnormal state occurs are ignored. Further refer to Figure 3Schematic diagram of the generation process of the filled blood oxygen signal. Among them, the abnormal blood oxygen signal sequence may include: abnormal blood oxygen signal A and abnormal blood oxygen signal B, that is, it represents that there are two local blood pressure signals in the blood pressure signal in an abnormal state. Therefore, the blood oxygen mean value of the normal blood oxygen signal between abnormal blood oxygen signal A and abnormal blood oxygen signal B can be used as the second interpolation between abnormal blood oxygen signal A and abnormal blood oxygen signal B, and the signal length of the normal blood oxygen signal between abnormal blood oxygen signal A and abnormal blood oxygen signal B can be used as the filling amount. And according to the second filling information, the blood oxygen signal is adjusted to obtain the filled blood oxygen signal. In this way, since the normal blood oxygen signals in the two abnormal blood oxygen signals are replaced with the second interpolation, for the part replaced with the second interpolation, the subsequent feature processing only needs to be calculated once and obtained by copying to get the feature values corresponding to the normal blood oxygen signal. At the same time, the time interval feature between every two abnormal blood oxygen signals can be retained through the setting of the filling amount. The generation methods of the first filling information and the third filling information are the same and will not be elaborated here.

[0030] The third step is to generate the filled blood pressure signal, the filled blood oxygen signal, and the filled heart rate signal according to the above abnormal blood pressure signal sequence, the above abnormal blood oxygen signal sequence, the above abnormal heart rate signal sequence, the above first filling information sequence, the above second filling information sequence, and the above third filling information sequence.

[0031] In practice, the first filling information is used to replace the signal values of the normal blood pressure signal between every two abnormal blood pressure signals. The second filling information is used to update the signal values of the normal blood oxygen signal between every two abnormal blood oxygen signals. The third filling information is used to update the signal values of the normal heart rate signal between every two abnormal heart rate signals.

[0032] The fourth step is to generate the above emotion state information according to the above filled blood pressure signal, the above filled blood oxygen signal, the above filled heart rate signal, and the pre-trained emotion state prediction model.

[0033] Among them, the above emotion state prediction model includes: a signal compression module, a signal feature extraction model, and an emotion state classifier. The signal feature extraction module includes: a signal time window splitter and a signal encoder. The signal compression module is used to perform signal compression on the filled blood pressure signal, the above filled blood oxygen signal, and the above filled heart rate signal at the same scale. The signal time window splitter is used to perform simultaneous window splitting on the compressed blood pressure signal, the compressed blood oxygen signal, and the compressed heart rate signal.

[0034] As an example, see Figure 4Schematic diagram of the generation process of the emotional state information shown, where the signal compression module will compress the filled blood pressure signal, the filled blood oxygen signal, and the filled heart rate signal. Specifically, since the sampling frequencies of the blood oxygen signal, the blood pressure signal, and the heart rate signal are different, there is a problem that they cannot be directly aligned subsequently, and at the same time, there is also a problem that the large amount of data increases the subsequent calculation pressure. Therefore, the signal compression module uses the method of downsampling with a variable sampling rate to compress the filled blood pressure signal, the filled blood oxygen signal, and the filled heart rate signal. Specifically, taking Figure 4 the filled blood oxygen signal shown as an example, the filled blood oxygen signal can be downsampled with a time interval T2 as the variable sampling rate to obtain a 1×M blood oxygen signal vector. Similarly, for the filled heart rate signal, the filled heart rate signal can be downsampled with a time interval T3 as the variable sampling rate. And for the filled blood pressure signal, the filled blood pressure signal can be downsampled with a time interval T1 as the variable sampling rate. Among them, the time interval T1, the time interval T2, and the time interval T3 are different. Further, since the blood pressure signal vector, the blood oxygen signal vector, and the heart rate signal vector processed by the signal compression module are all 1×M one-dimensional vectors. Therefore, they can be directly superimposed to form a 3×M feature map. Immediately afterwards, the signal time window splitter divides the 3×M feature map with a window size of 3×N and N as the step length to obtain M / N groups of 3×M feature matrices. Further, the signal encoder can encode the M / N groups of 3×M feature matrices according to the time sequence. In particular, the signal encoder adopts the Transformer Encoder structure included in the ViT model, and then obtains a L×H feature map. Finally, the emotional state classifier maps the L×H feature map to obtain a 1×K classification vector representing K emotional categories. In particular, the emotional state classifier adopts the MLP head structure. The above emotional state prediction model, as one of the invention points of the present disclosure, realizes signal alignment and emotion recognition under different signal sampling frequencies.

[0035] Step 5, generate an initial sleep physiological data feature map according to the above sleep physiological data.

[0036] In practice, sleep data is sampled at a granularity of days, that is, the scale of sleep data in the sleep data sequence is consistent in the time dimension. That is, a 1×T feature vector can be constructed for each sleep data. The length of T can be an hour granularity, or a finer granularity of half an hour. For example, T can be 24 or 48. Assume that the sleep data A includes a sleep duration of T1 and a wake duration of T2. The proportion of light sleep is R1% and the proportion of deep sleep is R2%. Therefore, taking T as 24 as an example, the vector length corresponding to the wake duration is 24 / T2. The vector length corresponding to the sleep duration is 24 / T1. The vector length corresponding to the proportion of light sleep is (24 / T1)×R1% and the vector length corresponding to the proportion of deep sleep is (24 / T1)×R2%. Next, the above-mentioned execution subject can mark the vector values in the 1×T vector in combination with the sleep distribution by means of a heat map. And multiple 1×T vectors are spliced along the time dimension to obtain the initial sleep physiological data feature map. The vector dimension of the initial sleep physiological data feature map is C×T, where C is the number of sleep data in the sleep data sequence.

[0037] In the sixth step, a graph feature extraction module is used to extract graph features from the initial sleep physiological data feature graph to generate a sleep feature heat map.

[0038] Among them, the heat map feature extraction module uses the feature pyramid network as the main structure. According to the length of T, the number of network layers of the heat map feature extraction module can be dynamically set. For example, when the value of T is small, a 3-layer feature pyramid network can be used. For another example, when the value of T is large, a 7-layer feature pyramid network can be used. In particular, the image size of the sleep feature heat map output by the heat map feature extraction modules of different depths is consistent.

[0039] The seventh step is to generate the sleep state information according to the sleep state classifier and the sleep feature heat map, wherein the image feature extraction module and the sleep state classifier are included in the sleep state recognition model.

[0040] The sleep state classifier uses three serially connected fully connected layers to reduce the dimension of the sleep feature heat map to 1×S, where S is the number of sleep state types.

[0041] Step 102 , in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, corresponding self-evaluation data of the postoperative patient is collected through a dynamically generated data collection scale.

[0042] In some embodiments, the above-mentioned execution entity may, in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collect the self-evaluation data corresponding to the postoperative patients through a dynamically generated data collection scale. Among them, the above-mentioned self-evaluation data represents the patients' self-evaluations from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression. The data collection scale includes multiple self-evaluation questions in the above 5 dimensions. The self-evaluation data is the answer results of the postoperative patients to the self-evaluation questions. In practice, the above-mentioned execution entity may combine the European Organization for Research and Treatment of Cancer Quality of Life Core Questionnaire, the Thyroid Cancer Specific Quality of Life Scale, the Perceived Stress Scale, the Pittsburgh Sleep Quality Index, and the Fear of Progression Simplified Scale, and dynamically construct a data collection scale by means of question extraction.

[0043] In practice, in the face of the challenge of the complexity of the recurrence mechanism, it is urgent to break through the traditional biomedical model and construct a multi-dimensional risk prevention and control system. Research shows that lifestyle adjustments may affect the tumor microenvironment by regulating chronic inflammation, oxidative stress and other pathways. Lifestyle-related factors such as negative emotions, psychological stress, and sleep disorders have been proven to be closely related to immune regulation functions and may indirectly affect the recurrence process. Therefore, integrating lifestyle and psychological interventions into the recurrence prevention and control system not only conforms to the concept of "moving the prevention and control forward" in tumor prevention and treatment, but also provides new ideas for establishing personalized prevention and control plans, realizing the leap from single medical intervention to the "biological-psychological-social" comprehensive management model.

[0044] In some optional implementation manners of some embodiments, the above-mentioned execution entity collects the self-evaluation data corresponding to the above-mentioned postoperative patients through a dynamically generated data collection scale, including: The first step is to obtain historical question sequence information.

[0045] Among them, the historical question sequence information represents the question order of the self-evaluation questions in the historically generated data collection scale.

[0046] The second step is to generate an initial data collection scale, where the above-mentioned initial data collection scale includes at least one self-evaluation question set from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression.

[0047] In practice, the above-mentioned execution entity may extract self-evaluation questions from the question bank containing self-evaluation questions in the European Organization for Research and Treatment of Cancer Quality of Life Core Questionnaire, the Thyroid Cancer Specific Quality of Life Scale, the Perceived Stress Scale, the Pittsburgh Sleep Quality Index, and the Fear of Progression Simplified Scale as the initial data collection scale.

[0048] In the third step, according to the above historical question sequence information, adjust the question sequence of at least one self-evaluation question included in the above initial data collection scale to obtain a candidate data collection scale.

[0049] In practice, for each self-evaluation question among at least one self-evaluation question included in the initial data collection scale, when the question order of the above self-evaluation question is the same as the question order of the self-evaluation question characterized in the historical question sequence information, adjust the question order of the self-evaluation question. The reason is that since the total number of self-evaluation questions corresponding to the scale is limited, after postoperative patients fill in at a certain frequency, especially when the question order remains unchanged, there may be a problem of forming a filling habit, rather than answering the self-evaluation questions based on actual feelings, thus affecting the accuracy of the self-evaluation data obtained from the filling. Therefore, it is necessary to adjust the question order of the self-evaluation questions in combination with the historical question sequence information to avoid postoperative patients relying on the filling order to answer.

[0050] In the fourth step, according to the above emotional state information and the above sleep state information, update the question weights of the self-evaluation questions corresponding to the psychological stress dimension in the above candidate data collection scale, and update the question weights of the self-evaluation questions corresponding to the sleep quality dimension to obtain the above data collection scale.

[0051] In practice, the initial weights of the self-evaluation questions in the candidate data collection scale are the same. Considering that the self-evaluation questions mainly rely on the self-feelings of postoperative patients to answer, there is a certain degree of subjectivity. Therefore, update the question weights of the self-evaluation questions corresponding to the psychological stress dimension and update the question weights of the self-evaluation questions corresponding to the sleep quality dimension. Specifically, the weight update depends on the emotional state information and the sleep state information. For the self-evaluation questions corresponding to the psychological stress dimension, when the filling result corresponding to the self-evaluation questions corresponding to the psychological stress dimension is opposite to the emotional state characterized by the emotional state information, reduce the question weight of the self-evaluation questions corresponding to the psychological stress dimension. When the filling result of the self-evaluation questions corresponding to the sleep quality dimension is opposite to the sleep state characterized by the sleep state information, reduce the question weight of the self-evaluation questions corresponding to the sleep quality dimension. If they are consistent, keep the question weight unchanged.

[0052] In the fifth step, send the above data collection scale to the first communication terminal associated with the above postoperative patient.

[0053] In practice, the above postoperative patient can answer the self-evaluation questions included in the data collection scale through the first communication terminal. Specifically, an online filling page can be generated in combination with the data collection scale for the postoperative patient to fill in and submit self-evaluation data.

[0054] In the sixth step, determine the filling result of the above data collection scale returned by the first communication terminal as the above self-evaluation data.

[0055] Step 103: Generate a recurrence prediction result according to the multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and the pre-trained recurrence prediction model for papillary thyroid carcinoma.

[0056] In practice, the above-mentioned execution entity can generate a recurrence prediction result according to the multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and the pre-trained recurrence prediction model for papillary thyroid carcinoma. Among them, the multimodal test data represents the test results obtained by the patient through the test items. Specifically, the multimodal examination and test data include, but are not limited to: ultrasonic images obtained by ultrasonic examination, index data related to puncture pathological examination, free triiodothyronine (FT3) index value, free thyroxine (FT4) index value, total triiodothyronine (TT3) index value, total thyroxine (TT4) index value, thyroid stimulating hormone (TSH), thyroid peroxidase antibody (TPOAb), and anti-thyroglobulin antibody (TGAb) index value. The above recurrence prediction result includes: recurrence risk type and recurrence risk confidence level. Among them, the recurrence prediction model for papillary thyroid carcinoma is trained by a supervised training method. Specifically, the multimodal test data and self-evaluation data corresponding to the collected recurrent patients can be used as training samples, and the recurrence risk type corresponding to the recurrent patients can be used as a sample label for model training.

[0057] Optionally, the papillary thyroid carcinoma recurrence prediction model includes: an examination test data feature extraction module, a self-evaluation data feature extraction module, a feature fusion module, a deep feature extraction module, and a recurrence risk type classifier. Among them, the test data feature extraction module includes: at least two sub-feature extraction modules. Among them, at least two sub-feature extraction modules correspond to different modality types. Specifically, at least two sub-feature extraction modules may include: sub-feature extraction module A and sub-feature extraction module B. Among them, sub-feature extraction module A is used to extract features from multi-modal diagnostic data of the image modality. Sub-feature extraction module B is used to extract features from multi-modal diagnostic data of the index value modality. Specifically, sub-feature extraction module A uses ResNet50 as the backbone network structure. Sub-feature extraction module B uses a convolutional neural network model as the backbone network structure. The self-evaluation data feature extraction module uses an encoder-decoder network based on the Transformer structure as the backbone network structure. Considering that the feature dimensions output by the test data feature extraction module and the self-evaluation data feature extraction module are inconsistent, the feature fusion module is used to unify the dimensions of the features output by sub-feature extraction module A, sub-feature extraction module B, and the self-evaluation data feature extraction module and splice them. Specifically, the horizontal dimension of the features is ensured to be consistent by padding with 0, and they are spliced vertically. The deep feature extraction module uses a feature pyramid network as the backbone network to extract features from multiple receptive fields. Finally, the recurrence risk type classifier uses a fully connected layer to downsample the features output by the feature pyramid network into a one-dimensional feature vector representing the recurrence prediction type.

[0058] In some optional implementation manners of some embodiments, generating a recurrence prediction result according to the multi-modal test data associated with the postoperative patient, the self-evaluation data, and the pre-trained papillary thyroid carcinoma recurrence prediction model includes: In the first step, through the above-mentioned test data feature extraction module, multi-modal feature extraction is performed on the above-mentioned multi-modal test data to generate a test data feature map.

[0059] In practice, at least two sub-feature extraction modules may include: sub-feature extraction module A and sub-feature extraction module B. Among them, sub-feature extraction module A is used to extract features from multi-modal diagnostic data of the image modality. Sub-feature extraction module B is used to extract features from multi-modal diagnostic data of the index value modality. Specifically, the above-mentioned execution subject takes the feature dimension of the features output by sub-feature extraction module A as a benchmark (for example, the dimension is M×N), maps the feature dimension of the features output by sub-feature extraction module B from, for example, 1×K to 1×N and then splices it with the features output by sub-feature extraction module A to obtain a test data feature map of (1 + M)×N.

[0060] In the second step, through the above self-evaluation data feature extraction module, semantic features of the above self-evaluation data are extracted to obtain a self-evaluation data feature map.

[0061] In practice, the self-evaluation data feature extraction module uses an encoder-decoder network based on the Transformer structure as the backbone network structure. The overall semantics of the answer results and the corresponding self-evaluation questions in the self-evaluation data are extracted, and multiple semantic features obtained are concatenated in the order of the self-evaluation questions in the data collection scale to obtain a self-evaluation data feature map.

[0062] In the third step, through the above feature fusion module, the above inspection data feature map and the above self-evaluation data feature map are fused to obtain a fused feature map.

[0063] In practice, feature fusion includes two parallel feature mapping networks, and the feature mapping network is composed of a convolutional neural network. Considering only the unification in the vector space, the feature mapping network adopts a relatively shallow network structure (for example, serially composed of 3 to 5 convolutional layers) to map the inspection data feature map and the above self-evaluation data feature map to the same feature space, where the horizontal dimension of the mapped inspection data feature map is the same as that of the mapped self-evaluation data feature map. Therefore, by vertically concatenating the mapped inspection data feature map and the mapped self-evaluation data feature map, a fused feature map is obtained.

[0064] In the fourth step, through the above deep feature extraction module, deep features of the above fused feature map are extracted to obtain a deep feature map.

[0065] The deep features of the fused features are extracted under different receptive fields through a deep feature extraction network with a feature pyramid network as the backbone network, and the feature maps under different receptive fields are superimposed to obtain a deep feature map.

[0066] In the sixth step, according to the above recurrence risk type classifier and the above deep feature map, the above recurrence prediction result is generated.

[0067] In practice, the deep feature map is mapped to a one-dimensional feature vector representing the recurrence prediction type through a recurrence risk type classifier.

[0068] As one of the invention points of the present disclosure, the above thyroid papillary carcinoma recurrence prediction model can achieve accurate and effective recurrence risk prediction by combining multi-modal inspection data and self-evaluation data.

[0069] Step 104, according to the trigger condition triggered by the recurrence prediction result, send a recurrence risk prompt associated with the trigger condition.

[0070] In some embodiments, the above-mentioned execution entity may, according to the triggering condition triggered by the recurrence prediction result, initiate a recurrence risk reminder associated with the triggering condition. In practice, by setting different levels of triggering conditions according to different recurrence risk types, different objects are associated and reminded accordingly.

[0071] In some optional implementation manners of some embodiments, the above-mentioned execution entity, according to the triggering condition triggered by the above-mentioned recurrence prediction result, initiates a recurrence risk reminder associated with the above-mentioned triggering condition, including: First step, in response to the above-mentioned recurrence prediction result triggering the first triggering condition, send a first recurrence risk reminder to the first communication terminal associated with the above-mentioned postoperative patient.

[0072] Among them, the above-mentioned first recurrence risk reminder is used to prompt the above-mentioned postoperative patient to have a postoperative follow-up visit. In practice, the first triggering condition is: the recurrence risk type included in the recurrence prediction result is a medium-low risk recurrence risk type and the recurrence risk confidence level is greater than the first preset risk confidence level.

[0073] Second step, in response to the above-mentioned recurrence prediction result triggering the second triggering condition, generate a recurrence risk report according to the above-mentioned physiological data, the above-mentioned self-evaluation data, and the above-mentioned recurrence prediction result.

[0074] In practice, the second triggering condition is: the recurrence risk type included in the recurrence prediction result is a high risk recurrence risk type and the recurrence risk confidence level is greater than the second preset risk confidence level.

[0075] Third step, push the above-mentioned recurrence risk report to the second communication terminal.

[0076] Among them, the above-mentioned second communication terminal is bound to the attending doctor corresponding to the above-mentioned postoperative patient.

[0077] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: Through the application of the recurrence prediction method for papillary thyroid carcinoma in postoperative patients in some embodiments of the present disclosure, the recurrence prediction for postoperative patients with papillary thyroid carcinoma is effectively realized. Specifically, first, according to the physiological data corresponding to the postoperative patients, the emotional state information and sleep state information for the above-mentioned postoperative patients are generated. Among them, the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the operation time. In practice, there is still a certain recurrence risk of papillary thyroid carcinoma after surgery. At the same time, the emotional state and sleep state of postoperative patients may also have a significant impact on thyroid function, thereby affecting the recurrence of the thyroid. Therefore, the present disclosure generally obtains the physiological data of postoperative patients, and then judges the emotional state and sleep state of postoperative patients. Secondly, in response to the above-mentioned emotional state information indicating emotional abnormality and / or the above-mentioned sleep state information indicating sleep state abnormality, through a dynamically generated data collection scale, the self-evaluation data corresponding to the above-mentioned postoperative patients is collected. Among them, the above-mentioned self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression. In practice, since the recurrence time varies among individuals, the method of regularly collecting data using a scale may delay the discovery of recurrence. At the same time, too high a data collection frequency may also cause resistance from patients. In addition, conventional scales often use a fixed format. Especially at a certain data collection frequency, patients are prone to the problem of habitual filling, resulting in inaccurate self-evaluation data collected. Therefore, the present disclosure determines the timing of data collection by combining the emotional state and sleep state, and dynamically generates the scale, so as to dynamically adjust the collection time and collection method according to individual circumstances. Then, according to the multimodal examination and test data associated with the above-mentioned postoperative patients, the above-mentioned self-evaluation data, and the pre-trained recurrence prediction model for papillary thyroid carcinoma, a recurrence prediction result is generated. The above-mentioned recurrence prediction result includes: recurrence risk type and recurrence risk confidence level. By combining self-evaluation data, examination and test data, and the prediction model, accurate prediction of recurrence is thus achieved. Finally, according to the trigger condition triggered by the above-mentioned recurrence prediction result, a recurrence risk reminder is sent to the initiator associated with the above-mentioned trigger condition. In this way, the risk reminder is automatically triggered, so that patients can seek medical treatment in time when they have a recurrence risk. Through this method, the recurrence prediction for postoperative patients with papillary thyroid carcinoma is effectively realized. Further reference Figure 5 , as an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a recurrence prediction device for papillary thyroid carcinoma in postoperative patients. These device embodiments correspond to Figure 1 the method embodiments shown, and the recurrence prediction device for papillary thyroid carcinoma in postoperative patients can be specifically applied to various electronic devices.

[0078] As Figure 5 shown, the papillary thyroid carcinoma recurrence prediction device 500 for postoperative patients in some embodiments includes: a first generation unit 501, a collection unit 502, a second generation unit 503, and a triggering unit 504. Among them, the first generation unit 501 is configured to generate emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient. Among them, the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data collection window corresponding to the physiological data is determined by the time span from the surgery time; the collection unit 502 is configured to respond to the emotional state information indicating emotional abnormality and / or the sleep state information indicating abnormal sleep state, and collect the self-evaluation data corresponding to the postoperative patient through a dynamically generated data collection scale, where the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; the second generation unit 503 is configured to generate a recurrence prediction result according to the multimodal examination data associated with the postoperative patient, the self-evaluation data, and a pre-trained papillary thyroid carcinoma recurrence prediction model, and the recurrence prediction result includes: a recurrence risk type and a recurrence risk confidence level; the triggering unit 504 is configured to trigger a recurrence risk prompt associated with the triggering condition according to the triggering condition triggered by the recurrence prediction result.

[0079] It can be understood that the units described in the papillary thyroid carcinoma recurrence prediction device 500 for postoperative patients correspond to the respective steps in the method described in the reference Figure 1 description. Therefore, the operations, features, and beneficial effects described above for the method also apply to the papillary thyroid carcinoma recurrence prediction device 500 for postoperative patients and the units included therein, and will not be repeated here. Next, refer to Figure 6 , which shows a schematic structural diagram of an electronic device (for example, a computing device) suitable for implementing some embodiments of the present disclosure. Figure 6 The electronic device shown is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure. As Figure 6As shown in the figure, the computer device includes a processor, a memory, and a network interface connected via a system bus. Among them, the memory may include a non-volatile storage medium and an internal memory. The non-volatile storage medium can store an operating system and computer programs. The computer programs include program instructions, which, when executed, can cause the processor to execute any of the above methods. The processor is used to provide computing and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the operation of the computer programs in the non-volatile storage medium. When the computer programs are executed by the processor, the processor can be caused to execute any of the above methods. The network interface is used for network communication, such as sending the assigned tasks, etc. Those skilled in the art can understand that Figure 6 the structure shown in the figure is only a block diagram of some structures related to the solution of the present disclosure, and does not constitute a limitation on the computer device to which the solution of the present disclosure is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0080] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0081] Among them, in one embodiment, the above-mentioned processor is used to run a computer program stored in a memory to implement the following steps: generating, according to physiological data corresponding to a postoperative patient, emotional state information and sleep state information for the postoperative patient, where the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time; in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collecting, through a dynamically generated data collection scale, self-evaluation data corresponding to the postoperative patient, where the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; generating a recurrence prediction result according to multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and a pre-trained recurrence prediction model for papillary thyroid carcinoma, where the recurrence prediction result includes: a recurrence risk type and a recurrence risk confidence level; and sending a recurrence risk prompt associated with the trigger condition according to a trigger condition triggered by the recurrence prediction result.

[0082] An embodiment of the present disclosure also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium, and program instructions are included in the computer program. The method implemented when the program instructions are executed can refer to various embodiments of the method for predicting the recurrence of papillary thyroid carcinoma in postoperative patients in the present disclosure.

[0083] Among them, the computer-readable storage medium may be an internal storage unit of the computer device in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a SmartMedia Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device.

[0084] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or system. Without further limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or system including the element.

[0085] The above description is only some preferred embodiments of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) disclosed in the embodiments of the present disclosure that have similar functions.

Claims

1. A method for predicting the recurrence of papillary thyroid carcinoma in postoperative patients, characterized in that, Including: Generating emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient, where the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the operation time; In response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collecting self-evaluation data corresponding to the postoperative patient through a dynamically generated data collection scale, where the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; Generating a recurrence prediction result according to the multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and a pre-trained recurrence prediction model for papillary thyroid carcinoma, where the recurrence prediction result includes: recurrence risk type and recurrence risk confidence level; Sending a recurrence risk reminder to an initiator associated with the trigger condition according to the trigger condition triggered by the recurrence prediction result.

2. The method according to claim 1, characterized in that The sending a recurrence risk reminder to an initiator associated with the trigger condition according to the trigger condition triggered by the recurrence prediction result includes: In response to the recurrence prediction result triggering a first trigger condition, sending a first recurrence risk reminder to a first communication terminal associated with the postoperative patient, where the first recurrence risk reminder is used to prompt the postoperative patient to have a postoperative follow-up visit; In response to the recurrence prediction result triggering a second trigger condition, generating a recurrence risk report according to the physiological data, the self-evaluation data, and the recurrence prediction result; Pushing the recurrence risk report to a second communication terminal, where the second communication terminal is bound to the attending doctor corresponding to the postoperative patient.

3. The method according to claim 2, wherein The basic physiological data includes: blood pressure signal, blood oxygen signal, heart rate signal, the sleep physiological data includes: a sleep data sequence, and the sleep data includes: sleep duration, wakefulness duration, light sleep ratio, deep sleep ratio, and sleep distribution; and The generating emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient includes: Performing abnormal signal localization on the blood pressure signal, the blood oxygen signal, and the heart rate signal respectively to obtain an abnormal blood pressure signal sequence, an abnormal blood oxygen signal sequence, and an abnormal heart rate signal sequence; Determining a first filling information sequence, a second filling information sequence, and a third filling information sequence according to the blood pressure signal, the blood oxygen signal, the heart rate signal, the abnormal blood pressure signal sequence, the abnormal blood oxygen signal sequence, and the abnormal heart rate signal sequence, where the first filling information includes: a first interpolation and a filling amount, the first interpolation represents the signal interpolation between two adjacent abnormal blood pressure signals, the second filling information includes: a second interpolation and a filling amount, the second interpolation represents the signal interpolation between two adjacent abnormal blood oxygen signals, and the third filling information includes: a third interpolation and a filling amount, the third interpolation represents the signal interpolation between two adjacent abnormal heart rate signals; Generate the filled blood pressure signal, filled blood oxygen signal, and filled heart rate signal according to the abnormal blood pressure signal sequence, the abnormal blood oxygen signal sequence, the abnormal heart rate signal sequence, the first filling information sequence, the second filling information sequence, and the third filling information sequence.

4. The method according to claim 3, wherein The generating of the emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient further includes: Generate the emotional state information according to the filled blood pressure signal, the filled blood oxygen signal, the filled heart rate signal, and a pre-trained emotional state prediction model, where the emotional state prediction model includes: a signal compression module, a signal feature extraction model, and an emotional state classifier. The signal feature extraction module includes: a signal time window splitter and a signal encoder. The signal compression module is used to perform signal compression on the filled blood pressure signal, the filled blood oxygen signal, and the filled heart rate signal at the same scale. The signal time window splitter is used to perform simultaneous window splitting on the compressed blood pressure signal, compressed blood oxygen signal, and compressed heart rate signal. Generate an initial sleep physiological data feature map according to the sleep physiological data. Perform graph feature extraction on the initial sleep physiological data feature map through a graph feature extraction module to generate a sleep feature heat map. Generate the sleep state information according to a sleep state classifier and the sleep feature heat map, where the graph feature extraction module and the sleep state classifier are included in a sleep state recognition model.

5. The method according to claim 4, wherein The collecting of the self-evaluation data corresponding to the postoperative patient through a dynamically generated data collection scale includes: Obtain historical question order information, where the historical question order information represents the question order of the self-evaluation questions in the historically generated data collection scale. Generate an initial data collection scale, where the initial data collection scale includes at least one self-evaluation question set under the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression. Adjust the question order of at least one self-evaluation question included in the initial data collection scale according to the historical question order information to obtain a candidate data collection scale. Update the question weights of the self-evaluation questions corresponding to the psychological stress dimension and the self-evaluation questions corresponding to the sleep quality dimension in the candidate data collection scale according to the emotional state information and the sleep state information to obtain the data collection scale. Send the data collection scale to a first communication terminal associated with the postoperative patient. Determine the filling result for the data collection scale returned by the first communication terminal as the self-evaluation data.

6. The method according to claim 5, wherein The papillary thyroid carcinoma recurrence prediction model includes: an examination and test data feature extraction module, a self-evaluation data feature extraction module, a feature fusion module, a deep feature extraction module, and a recurrence risk type classifier, where the examination and test data feature extraction module includes: at least two sub-feature extraction modules; and Generating a recurrence prediction result according to the multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and a pre-trained recurrence prediction model for papillary thyroid carcinoma, including: Through the examination and test data feature extraction module, performing multimodal feature extraction on the multimodal examination and test data to generate an examination and test data feature map; Through the self-evaluation data feature extraction module, performing semantic feature extraction on the self-evaluation data to obtain a self-evaluation data feature map; Through the feature fusion module, fusing the examination and test data feature map and the self-evaluation data feature map to obtain a fused feature map; Through the deep feature extraction module, performing deep feature extraction on the fused feature map to obtain a deep feature map; Generating the recurrence prediction result according to the recurrence risk type classifier and the deep feature map.

7. A recurrence prediction device for papillary thyroid carcinoma in postoperative patients, characterized in that, Including: A first generation unit configured to generate emotional state information and sleep state information for the postoperative patient according to the physiological data corresponding to the postoperative patient, where the physiological data includes: basic physiological data and sleep physiological data, and the window length of the data acquisition window corresponding to the physiological data is determined by the time span from the surgery time; An acquisition unit configured to, in response to the emotional state information indicating abnormal emotions and / or the sleep state information indicating abnormal sleep states, collect the self-evaluation data corresponding to the postoperative patient through a dynamically generated data collection scale, where the self-evaluation data represents the patient's self-evaluation from the dimensions of quality of life, thyroid specificity, psychological stress, sleep quality, and fear of disease progression; A second generation unit configured to generate a recurrence prediction result according to the multimodal examination and test data associated with the postoperative patient, the self-evaluation data, and a pre-trained recurrence prediction model for papillary thyroid carcinoma, where the recurrence prediction result includes: a recurrence risk type and a recurrence risk confidence level; An initiation unit configured to initiate a recurrence risk reminder associated with the trigger condition according to the trigger condition triggered by the recurrence prediction result.

8. An electronic device, characterized in that, Including: One or more processors; A storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer-readable medium, characterized in that, Having a computer program stored thereon, where the computer program, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images