CN-122000050-A - Lifetime prediction method and device, electronic equipment and storage medium

Abstract

The application discloses a lifetime prediction method, a lifetime prediction device, electronic equipment and a storage medium. The method comprises the steps of extracting image features of medical images of a patient by adopting an image analysis agent to generate image index vectors, extracting clinical semantic features of unstructured clinical texts and structured laboratory indexes of the patient by adopting a clinical information processing agent to generate clinical text semantic vectors and laboratory semantic vectors, and carrying out hierarchical fusion processing on the image index vectors, the clinical text semantic vectors and the laboratory semantic vectors by adopting a fusion reasoning agent to generate life cycle prediction results of the patient. The application predicts the lifetime based on the multi-mode medical data, improves the prediction effect, adopts a distributed multi-agent cooperative architecture, ensures that the output prediction result has high interpretability, and enables doctors to know the prediction made by the system based on what clinical indexes (such as tumor size, form and the like).

Inventors

• LI TAO
• XU CHENGLIN

Assignees

• 麒麟合盛网络技术股份有限公司

Dates

Publication Date

20260508

Application Date

20260115

Claims (10)

1. 1. A method of lifetime prediction, comprising: image analysis intelligent body is adopted to extract image characteristics of the medical image of the patient so as to generate an image index vector; Respectively extracting clinical semantic features of unstructured clinical texts and structured laboratory indexes of the patient by adopting a clinical information processing agent to generate clinical text semantic vectors and laboratory semantic vectors; And carrying out hierarchical fusion processing on the image index vector, the clinical text semantic vector and the laboratory semantic vector by adopting a fusion reasoning agent so as to generate a life cycle prediction result of the patient.
2. 2. The method of claim 1, wherein performing image feature extraction on the medical image of the patient to generate an image index vector comprises: performing focus segmentation on the medical image through a deep neural network model so as to locate a focus; calculating the image index score of the focus by adopting a space geometric algorithm; and determining the image index score as the image index vector.
3. 3. The method of claim 2, wherein the image index score comprises a score She Zheng, the computing the image index score for the lesion using a spatial geometry algorithm comprising: Constructing a three-dimensional surface grid model of the focus, and calculating the average curvature of each vertex in the grid model; determining the score She Zheng according to the frequency and depth of the positive and negative alternation of the average curvature.
4. 4. The method of claim 2, wherein the image index score comprises a spur score, the computing the image index score for the lesion using a spatial geometry algorithm comprising: Taking the boundary of the focus as a starting point, and detecting rays along the normal direction; Calculating gray gradient changes outside the boundary; Counting the number of radial protrusions with length exceeding a length threshold and aspect ratio greater than a proportion threshold; And determining the burr score according to the gray gradient change and the number of the radial protrusions.
5. 5. The method of claim 2, wherein the image index score comprises a pleural stretch score, the calculating the image index score for the lesion using a spatial geometry algorithm comprising: identifying a pleural line boundary nearest the lesion; Establishing a region of interest between the edge of the lesion and the nearest pleural line boundary, and detecting bridging structures within the region of interest; Calculating a displacement vector of the pleura at the connection of the bridging structure, and determining an inner angle of the pleural cavity according to the displacement vector; Determining the pleural traction score based on the thickness of the bridging structure and the interior angle.
6. 6. The method of claim 1, wherein extracting clinical semantic features from the clinical text using a clinical information processing agent to generate the clinical text semantic vector comprises: identifying a plurality of medical entities in the clinical text; performing standard semantic anchor point alignment processing on the plurality of medical entities; and quantifying the plurality of aligned medical entities to obtain the clinical text semantic vector.
7. 7. The method of claim 1, wherein the hierarchically fusing the image index vector, the clinical text semantic vector, and the laboratory semantic vector with a fusion inference agent to generate a survival prediction result for the patient comprises: Inputting the clinical text semantic vector and the laboratory semantic vector into a first-level attention mechanism network for fusion to obtain a clinical fusion semantic vector; Inquiring in the clinical fusion semantic vector according to the image index vector to generate a target fusion feature vector; And generating the lifetime prediction result according to the image index vector, the target fusion feature vector and the weights, wherein the weights are automatically adjusted according to the data quality and/or the absence of the corresponding vector.
8. 8. A lifetime prediction device, comprising: The image analysis intelligent body is used for extracting image features of the medical image of the patient so as to generate an image index vector; The clinical information processing intelligent body is used for respectively extracting clinical semantic features of unstructured clinical texts and structured laboratory indexes of the patient so as to generate clinical text semantic vectors and laboratory semantic vectors; And the fusion reasoning intelligent agent is used for carrying out hierarchical fusion processing on the image index vector, the clinical text semantic vector and the laboratory semantic vector so as to generate a life cycle prediction result of the patient.
9. 9. An electronic device comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, which program or instruction when executed by the processor implements the steps of the method of any of claims 1-8.
10. 10. A readable storage medium, characterized in that it stores thereon a program or instructions, which when executed by a processor, implement the steps of the method according to any of claims 1-8.

Description

Lifetime prediction method and device, electronic equipment and storage medium Technical Field The application belongs to the technical field of artificial intelligence, and particularly relates to a lifetime prediction method, a lifetime prediction device, electronic equipment and a storage medium. Background The life-time prediction of patients with malignant tumors (such as lung cancer, liver cancer and the like) has important significance for clinical decision-making, scheme making and resource allocation. In the related art, as shown in fig. 1, raw image data or clinical medical record data is input into a black box-like end-to-end deep neural network to obtain a single survival probability value. However, the above scheme focuses on single-mode medical data (such as only for images or only for medical records), the prediction effect is poor, in addition, the deep neural network algorithm logic is black-boxed, the prediction result is directly output, the interpretation is lacking, and a doctor cannot know the prediction made by the system based on what clinical indexes (such as tumor size, morphology and the like). Disclosure of Invention The embodiment of the application aims to provide a lifetime prediction method, a lifetime prediction device, electronic equipment and a storage medium, which are used for solving the problems that in the related technology, single mode is focused, the prediction effect is poor, the algorithm logic is black-boxed, and doctors cannot know the clinical indexes (such as tumor size, tumor morphology and the like) of the system to make predictions. In order to achieve the above purpose, the embodiment of the present application adopts the following technical scheme: In a first aspect, an embodiment of the present application provides a lifetime prediction method, including performing image feature extraction on a medical image of a patient by using an image analysis agent to generate an image index vector, performing clinical semantic feature extraction on unstructured clinical text and structured laboratory indexes of the patient by using a clinical information processing agent to generate a clinical text semantic vector and a laboratory semantic vector, and performing hierarchical fusion processing on the image index vector, the clinical text semantic vector and the laboratory semantic vector by using a fusion inference agent to generate a lifetime prediction result of the patient. In a second aspect, the embodiment of the application provides a lifetime prediction device, which comprises an image analysis agent, a clinical information processing agent and a fusion reasoning agent, wherein the image analysis agent is used for extracting image features of a medical image of a patient to generate an image index vector, the clinical information processing agent is used for extracting clinical semantic features of unstructured clinical texts and structured laboratory indexes of the patient to generate clinical text semantic vectors and laboratory semantic vectors respectively, and the fusion reasoning agent is used for carrying out hierarchical fusion processing on the image index vector, the clinical text semantic vectors and the laboratory semantic vectors to generate lifetime prediction results of the patient. In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor, the program or instruction implementing the steps of the method according to the embodiment of the first aspect of the present application when executed by the processor. In a fourth aspect, embodiments of the present application provide a readable storage medium having stored thereon a program or instructions which when executed by a processor perform the steps of the method according to the embodiments of the first aspect of the present application. The above at least one technical scheme adopted by the embodiment of the application can achieve the following beneficial effects: The embodiment of the application predicts the lifetime based on the multi-mode medical data (medical images, unstructured clinical texts and structured laboratory indexes), and improves the prediction effect. The distributed multi-agent collaborative architecture is adopted, quantized characteristic vectors (image index vectors, clinical text semantic vectors and laboratory semantic vectors) are output in the process, so that the algorithm logic is visualized, the output prediction result has high interpretability, a doctor can know the prediction made by the system based on clinical indexes (such as tumor size, morphology and the like), and the user experience is improved. Drawings The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the applica