Construction method of noninvasive differential diagnosis model for tuberculous and cancerous pleural effusion

Abstract

The invention discloses a construction method of a tuberculous and cancerous pleural effusion non-invasive differential diagnosis model, and belongs to the technical field of medicine, and the method comprises the following steps: S1, constructing a deep learning model; s2, constructing a second-class deep learning model; s3, constructing three types of deep learning models; s4, constructing a four-class deep learning model; s5, constructing a five-class deep learning model; according to the method, a traditional Unet is taken as a basic skeleton, a deep learning model is constructed by adding a three-dimensional attention mechanism, the segmentation performance of the model is trained through a cut local CT image, the classification performance of the model is trained through an overall CT image and the segmented local CT image, and finally the diagnosis model is obtained. The constructed diagnosis model can realize differential diagnosis only through CT images during diagnosis, has the advantages of low cost, non-invasiveness and high accuracy, reduces the risk of complications of patients, reduces the medical cost, improves the diagnosis efficiency, and effectively relieves the pressure of clinical workers.

Description

technical field [0001] The invention belongs to the technical field of medicine, in particular to a method for constructing a non-invasive differential diagnosis model of tuberculous and cancerous pleural effusion. Background technique [0002] The identification of benign and malignant pleural effusion is a common focus and difficulty in clinical practice, among which the identification of tuberculous and lung cancer pleural effusion is very important. The diagnosis is based on pleural effusion, clinical signs, and the diagnostic accuracy is low. Therefore, a variety of examinations are needed to assist the diagnosis, such as routine pleural effusion and biochemical examination methods. However, thoracentesis exposes patients to a certain risk of complications, and the diagnostic efficiency is slow. Therefore, it is necessary to design a method for constructing a non-invasive differential diagnosis model of tuberculous and cancerous pleural effusion, and to solve the above ...

AI Technical Summary

Problems solved by technology

[0002] The identification of benign and malignant pleural effusions is a common focus and difficulty in clinical practice, among which the identification of tuberculous and lung cancer pleural effusion is very important. Diagnosis is based on clinical signs, and the diagnostic accuracy is low. Therefore, a variety of examinations are needed to assist the diagnosis, such as routine pleural effusion, biochemical examinations, and thoracentesis is an important measure to extract pleural effusion for routine biochemical examinations. However, thoracentesis has a certain risk of complications for patients, and the diagnostic efficiency is slow. Therefore, it is necessary to design a method for constructing a non-invasive differential diagnosis model for tuberculous and cancerous pleural effusions, and to solve the above problems through the non-invasive differential diagnosis model.

Images