New coronal pneumonia federal learning detection method based on DF

Abstract

The invention provides a new coronal pneumonia federal learning detection method based on DF (Dynamic Fusion). Since the X-Ray and lung CT images are of great significance to the diagnosis of the newcoronal pneumonia, and the federal learning can perform machine learning modeling on the premise of protecting the data privacy, a method for diagnosing the new coronal pneumonia by using the federallearning for the image data is designed. Data of different hospitals can be comprehensively utilized to establish a detection model based on deep learning. In order to reduce communication consumptionin federal learning, DF (Dynamic Fusion) is used for screening models waiting for uploading, whether the models are uploaded and fused or not is determined according to the accuracy of a local modeltrained by each client node on a test set, and a Waiting Time is set by evaluating the computing power of each client node so as to effectively solve the waiting problem of a central node. According to the invention, the new coronal pneumonia can be accurately diagnosed according to the lung image data, and the communication loss is effectively reduced.

Description

technical field [0001] The present invention relates to the fields of federated learning and deep learning, in particular to a novel coronavirus pneumonia detection method based on federated learning and deep learning. Background technique [0002] Lung X-Ray and CT image data are of great significance to the diagnosis of COVID-19, and are an important link in the diagnosis of COVID-19. However, the examination data of each patient is as many as thousands of images, and the doctor's diagnosis alone requires a very high level of physician experience. High and very slow. Federated learning combines deep learning to process image data, and can comprehensively utilize data to quickly analyze images and make accurate diagnoses. The techniques closest to the present invention are: [0003] (1) Image classification algorithm based on deep learning: Image classification is to distinguish different types of images according to the semantic information of images, which is an importa...

AI Technical Summary

Problems solved by technology

However, this method only targets X-Ray data and does not fuse CT data, which is prone to misdiagnosis

Images