Search

CN-122025129-A - Grading determination method, device and equipment for tumor radiotherapy complications

CN122025129ACN 122025129 ACN122025129 ACN 122025129ACN-122025129-A

Abstract

The application provides a grading determination method, device and equipment for tumor radiotherapy complications, which can be applied to medical care scenes and comprise the steps that a cloud server acquires historical tumor radiotherapy data and multi-mode data from a database, the multi-mode data comprise structured data and unstructured data, the structured data comprise total planned absorbed dose, radiotherapy times, planned absorbed dose of radiotherapy each time and chemotherapy drugs, the unstructured data comprise self-timer images of patients and complications description texts provided by the patients, the cloud server determines a first grading value of the tumor radiotherapy complications of the patients based on the historical tumor radiotherapy data, the cloud server determines an adjustment factor of the first grading value based on the multi-mode data, and the cloud server corrects the first grading value based on the adjustment factor to obtain a second grading value of the tumor radiotherapy complications.

Inventors

  • ZHANG JINGJING
  • Che Suiyan
  • LIU MIN
  • XIA WEI

Assignees

  • 西安国际医学中心有限公司

Dates

Publication Date
20260512
Application Date
20260128

Claims (15)

  1. 1. A method for hierarchical determination of tumor radiotherapy complications, comprising: The cloud server acquires historical tumor radiotherapy data and multi-mode data from a database, wherein the multi-mode data comprises structured data and unstructured data, the structured data comprises total planned absorbed dose, radiotherapy times, planned absorbed dose of radiotherapy each time and chemotherapeutic drugs, and the unstructured data comprises a patient self-timer image and a complication description text provided by the patient; The cloud server determines a first hierarchical value of tumor radiotherapy complications of the patient based on the historical tumor radiotherapy data; the cloud server determines an adjustment factor of the first hierarchical value based on the multi-modal data; and the cloud server corrects the first grading value based on the adjustment factor to obtain a second grading value of the tumor radiotherapy complications.
  2. 2. The method of claim 1, wherein the cloud server determining a first hierarchical value of tumor radiotherapy complications for the patient based on the historical tumor radiotherapy data comprises: The cloud server determines a plurality of historical occurrence probabilities under a risk factor set of the patient and a plurality of proportions under the risk factor set based on the historical tumor radiotherapy data, wherein the risk factor set comprises a set formed by one or more risk factors inducing the patient to present the tumor radiotherapy complications, the plurality of historical occurrence probabilities are in one-to-one correspondence with a plurality of grading values of the tumor radiotherapy complications, and the plurality of proportions are in one-to-one correspondence with the plurality of grading values; And the cloud server determines the first grading value according to the historical occurrence probability and the proportions.
  3. 3. The method of claim 2, wherein the cloud server determining the first ranking value based on the historical occurrence probability and the plurality of proportions comprises: The cloud server determines a plurality of probabilities corresponding to the plurality of grading values one by one according to the following formula: ; Wherein the said Representing the probability of the patient developing a tumor radiotherapy complication with a grade value of i, the Representing the historical occurrence probability of tumor radiotherapy complications with a grading value of i under the risk factor set, wherein the tumor radiotherapy complications are obtained by the method Representing the proportion of tumor radiotherapy complications with a grading value of i under the risk factor set, wherein Representing the number of risk factors in the set of risk factors, ; And the cloud server determines a grading value corresponding to the maximum probability in the probabilities in the grading values as the first grading value.
  4. 4. The method of claim 1, wherein the cloud server determining the adjustment factor for the first hierarchical value based on the multimodal data comprises: the cloud server processes the structured data and obtains a first feature vector; the cloud server processes the self-timer image and obtains a second feature vector; the cloud server processes the complication description text and obtains a third feature vector; The cloud server obtains a fourth feature vector by splicing the first feature vector, the second feature vector and the third feature vector, and outputs the adjustment factor by using a regression network with the fourth feature vector as input, wherein the regression network comprises at least one full-connection layer and a Sigmoid activation function layer connected with the at least one full-connection layer, parameters of the at least one full-connection layer are pre-trained parameters, the number of nodes in the last full-connection layer connected with the Sigmoid activation function layer in the at least one full-connection layer is 1, and the value range of the adjustment factor is [0, 1].
  5. 5. The method of claim 4, wherein the cloud server processing the structured data and obtaining a first feature vector comprises: The cloud server normalizes the total planned absorbed dose, the radiotherapy times and the planned absorbed dose of each radiotherapy to obtain a normalized value of the total planned absorbed dose, a normalized value of the radiotherapy times and a normalized value of the planned absorbed dose of each radiotherapy; the cloud server multiplies the radiotherapy times by a radiotherapy period to obtain total treatment days, and performs normalization treatment on the total treatment days to obtain a normalization treatment value of the total treatment days; The cloud server utilizes the planned absorbed dose and the planned absorbed dose of each radiotherapy Sum divided by reference dose Obtaining an equivalent damage of the planned absorbed dose of each radiotherapy, multiplying the total planned absorbed dose by the equivalent damage to obtain an equivalent total dose, and normalizing the equivalent total dose to obtain a normalized value of the equivalent total dose, wherein, Represents the probability that a single ionization event will cause a double strand break or irreparable damage to the deoxyribonucleic acid DNA, Representing the probability that multiple independent ionization events together cause DNA double strand breaks or irreparable damage in the same region and the same repair time window; And the cloud server splices the normalized treatment value of the total planned absorbed dose, the normalized treatment value of the radiotherapy times, the normalized treatment value of the planned absorbed dose of each radiotherapy, the normalized treatment value of the total treatment days and the normalized treatment value of the equivalent total dose to obtain the first feature vector.
  6. 6. The method of claim 4, wherein the cloud server processes and obtains a second feature vector of the self-captured image, comprising: When the resolution of the self-timer image is larger than or equal to the preset resolution, the cloud server processes the self-timer image into a gray scale heat map with the same size as the self-timer image, wherein pixels in the gray scale heat map represent the probability of the existence of a focus; the cloud server performs mask processing on the gray scale heat image based on a predefined value and obtains a mask image, wherein pixels in the mask image represent whether a focus exists; the cloud server selects a local area with the area of a focus larger than or equal to a preset area in the mask image; and the cloud server performs feature extraction on the features of the self-shot image and the local area respectively to obtain a feature vector of the self-shot image and a feature vector of the local area, and obtains the second feature vector by fusing the feature vector of the self-shot image and the feature vector of the local area.
  7. 7. The method of claim 4, wherein the cloud server processing the complication description text and obtaining a third feature vector comprises: The cloud server performs word segmentation processing on the complication description text to obtain M word entities in the complication description text; The cloud server takes a query statement as input, invokes a language model to output M type tags corresponding to the M word entities one by one, wherein the query statement is used for requesting to query the M type tags, the type tags are used for indicating the types of the word entities, and the types of the word entities are any one of the word entity types used for describing the complication part, the word entity types used for describing the symptom of the complication, the word entity types used for describing the severity of the complication and the invalid types; The cloud server deletes word entities with invalid word entity types in the M word entities based on the M type tags to obtain N word entities, wherein M > N is more than or equal to 1; The cloud server searches the number of each word entity in the dictionary corresponding to each word entity in the N word entities based on N dictionaries corresponding to the N word entities one by one to obtain N numbers corresponding to the N word entities one by one, and determines N word vectors corresponding to the N word entities one by one based on the N numbers; And multiplying or adding the N word vectors by the cloud server to obtain the third feature vector.
  8. 8. The method of claim 1, wherein the adjustment factor is used to characterize a probability of increasing the first hierarchical value, wherein the cloud server corrects the first hierarchical value based on the adjustment factor to obtain a second hierarchical value of the tumor radiotherapy complications, comprising: the cloud server determines the first classification value as the second classification value when the adjustment factor is smaller than the first threshold and the adjustment factor is larger than the second threshold, wherein the first threshold is larger than the second threshold, the first threshold is smaller than 1, and the second threshold is larger than 0; When the adjustment factor is greater than or equal to the first threshold value and the first grading value is smaller than a maximum grading value, the cloud server determines the sum of the first grading value and 1 as the second grading value; And under the condition that the adjustment factor is smaller than or equal to the second threshold value and the first grading value is larger than the minimum grading value, the cloud service determines the difference value between the first grading value and 1 as the second grading value.
  9. 9. The method according to any one of claims 1 to 8, further comprising: the cloud server determines a plurality of nursing actions connected with the second grading value in a knowledge graph based on the second grading value; The cloud server takes a grading feature vector obtained by the second grading value, an allergy history feature vector obtained by the allergy history of the patient and a radiotherapy position feature vector obtained by the radiotherapy position of the patient as inputs, and outputs a first score vector by using a strategy network, wherein the dimension of the first score vector is equal to the number of the plurality of nursing actions, and the first score vector comprises scores when the plurality of nursing actions are taken as recommended nursing actions; the cloud server determines a second score vector based on the first score vector, a knowledge map confidence coefficient vector, an allergy Shi Yanma vector obtained by mapping the allergy history of the patient to the space where the plurality of nursing actions are located, and a position mask vector obtained by mapping the nursing positions of the patient to the space where the plurality of nursing actions are located, wherein the knowledge map confidence coefficient is the confidence coefficient of a connecting edge between the second classification value and the plurality of nursing actions in the knowledge map; The cloud server takes the second score vector as input, activates a function layer to output a probability vector by using Softmax, and determines a nursing action corresponding to the maximum probability in the probability vector in the plurality of nursing actions as a recommended nursing action; and the cloud server sends information for indicating the recommended nursing action to nurse terminal equipment.
  10. 10. The method of claim 9, wherein the second classification value is a classification node in the knowledge-graph, the attribute of the classification node comprising a label for indicating a type of complication, the type of complication comprising oral mucositis or radiation skin injury.
  11. 11. The method of claim 9, wherein the cloud server determining a second score vector based on the first score vector, a knowledge-graph confidence vector, an allergy Shi Yanma vector resulting from mapping the allergy history of the patient to the space in which the plurality of care actions are located, and a site mask vector resulting from mapping the care location of the patient to the space in which the plurality of care actions are located, comprising: The cloud server adds the first score vector, the knowledge-graph confidence vector, the allergy Shi Yanma vector and the part mask vector to obtain the second score vector; The value of the first element in the allergy Shi Yanma vector and the value of the second element in the part mask vector are negative infinity, the nursing action corresponding to the first element is used as the medicine which is matched with the allergy history of the patient, and the nursing action corresponding to the second element is not matched with the nursing part of the patient.
  12. 12. The method according to claim 9, wherein the method further comprises: the cloud server receives the pain record value of the patient from the nurse terminal and calculates a reward value based on the pain record value; When the accumulated rewards reach X, subtracting the average rewards of the X rewards from each rewards of the X rewards by the cloud server to obtain X dominant values, and calculating the average value of the X dominant values, wherein X is more than or equal to 1; The cloud server determines the gradient of the strategy network based on the average value of the X dominant values; and the cloud server multiplies the gradient vector by a learning rate to obtain an updating coefficient, and updates parameters of the strategy network based on the updating coefficient.
  13. 13. A grade determining device for tumor radiotherapy complications, comprising: The system comprises a receiving module, a processing module and a processing module, wherein the receiving module is used for acquiring historical tumor radiotherapy data and multi-mode data from a database, the multi-mode data comprises structured data and unstructured data, the structured data comprises total planned absorbed dose, radiotherapy times, planned absorbed dose of radiotherapy each time and chemotherapeutic drugs, and the unstructured data comprises a patient self-timer image and a complication description text provided by the patient; A processing module for: determining a first grade value for tumor radiotherapy complications of the patient based on the historical tumor radiotherapy data; determining an adjustment factor for the first hierarchical value based on the multimodal data; and correcting the first grading value based on the adjustment factor to obtain a second grading value of the tumor radiotherapy complications.
  14. 14. An electronic device, comprising: a processor adapted to execute a computer program; a computer readable storage medium having stored therein a computer program which, when executed by the processor, implements the method of any one of claims 1 to 12.
  15. 15. A computer readable storage medium for storing a computer program which, when run on a computer, causes the computer to perform the method of any one of claims 1 to 12.

Description

Grading determination method, device and equipment for tumor radiotherapy complications Technical Field The application relates to the field of nursing of tumor radiotherapy complications in the field of medical nursing, and in particular relates to a grading determination method, device and equipment of tumor radiotherapy complications. Background In the related art, after a tumor patient is subjected to radiotherapy, if a tumor radiotherapy complication occurs, a nurse needs to determine a grading value according to a world health organization (World Health Organization, WHO) oral mucositis grading standard or a radiotherapy tumor cooperative group (Radiation Therapy Oncology Group, RTOG) radiotherapy skin toxicity grading standard prescribed by the Chinese nursing institute based on specific symptoms of the patient, and then care the patient according to a nursing scheme corresponding to the grading value prescribed by the Chinese nursing institute. However, tumor patients often experience tumor radiotherapy complications after radiotherapy, and if the grading value is determined according to the rules of the Chinese Care society, the phenomenon of "sheep-mending" occurs, namely, if the occurrence of tumor radiotherapy complications is known, the grading value is determined according to the specific symptoms of the tumor radiotherapy complications after the occurrence of the tumor radiotherapy complications, which can lead to excessively low nursing initiative and timeliness. Disclosure of Invention The embodiment of the application provides a grading determination method, device and equipment for tumor radiotherapy complications, which can improve the initiative and timeliness of nursing by a prediction and corrected grading value determination mode. The embodiment of the application provides a grading determination method of tumor radiotherapy complications, which comprises the steps that a cloud server obtains historical tumor radiotherapy data and multi-mode data from a database, the multi-mode data comprise structured data and unstructured data, the structured data comprise total planned absorbed dose, radiotherapy times, planned absorbed dose of radiotherapy each time and chemotherapy drugs, the unstructured data comprise self-timer images of patients and complications description texts provided by the patients, the cloud server determines a first grading value of the tumor radiotherapy complications of the patients based on the historical tumor radiotherapy data, the cloud server determines an adjustment factor of the first grading value based on the multi-mode data, and the cloud server corrects the first grading value based on the adjustment factor to obtain a second grading value of the tumor radiotherapy complications. In a second aspect, the embodiment of the application provides a grading determination device for tumor radiotherapy complications, which comprises a receiving module, a processing module and a grading determination module, wherein the receiving module is used for acquiring historical tumor radiotherapy data and multi-mode data from a database, the multi-mode data comprises structured data and unstructured data, the structured data comprises total planned absorbed dose, radiotherapy times, planned absorbed dose of radiotherapy each time and chemotherapy drugs, the unstructured data comprises a patient self-timer image and complications description text provided by the patient, the processing module is used for determining a first grading value of the tumor radiotherapy complications of the patient based on the historical tumor radiotherapy data, determining an adjustment factor of the first grading value based on the multi-mode data, and correcting the first grading value based on the adjustment factor to obtain a second grading value of the tumor radiotherapy complications. In a third aspect, an embodiment of the application provides an electronic device comprising a processor adapted to implement computer instructions, and a computer-readable storage medium storing computer instructions adapted to be loaded by the processor and to perform the method of the first aspect referred to above. In a fourth aspect, embodiments of the present application provide a computer-readable storage medium storing computer instructions that, when read and executed by a processor of a computer device, cause the computer device to perform the method of the first aspect referred to above. In a fifth aspect, embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, the processor executes the computer instructions, causing the computer device to perform the method of the first aspect referred to above. In the embodiment of the application, the cloud server determines a first gr