Neural network dynamic fusion method for genetic metabolic disease multi-center screening

Abstract

The invention discloses a genetic metabolic disease multi-center screening-oriented neural network dynamic fusion method. In the method, besides a plurality of screening centers, two types of nodes, namely task nodes and computing nodes, need to be configured, and the task nodes are responsible for management, distribution and maintenance of multi-center screening tasks; each screening center needs to be equipped with a computing node which is responsible for computing of a joint modeling task issued by a task node. The method disclosed by the invention is oriented to a hereditary metabolic disease multi-center screening scene, and fills the vacancy of a multi-center joint modeling method. Secondly, in combination with the characteristics of multiple genetic metabolic disease multi-center screening modeling tasks and large load pressure of neural network converged communication, the method adopts detective parameter sampling, evaluates the synchronization degree of multiple computational node model iteration, dynamically adjusts time nodes of model fusion, improves the fusion efficiency, reduces the communication frequency, and can effectively reduce the communication load of the whole task.

Description

technical field [0001] The invention belongs to the technical field of multi-center screening model construction, and relates to a neural network dynamic fusion method, in particular to a neural network dynamic fusion method for multi-center screening of genetic metabolic diseases. Background technique [0002] In recent years, the screening mode of genetic metabolic diseases has changed from the original single-center, single-hospital closed independent screening to multi-center collaborative joint screening, and various alliance forms such as medical alliances and medical communities have emerged , such as the National Children's Health and Disease Clinical Medical Research Center, the National Children's Regional Medical Center, etc., have strengthened the resource allocation, accuracy and efficiency of genetic metabolic disease screening. At the same time, the development of artificial intelligence technology has brought changes to the medical industry and proposed a new...

AI Technical Summary

Problems solved by technology

[0003] At present, some hospitals and screening centers have deployed artificial intelligence-assisted diagnosis platforms as pilots to improve the screening quality of the hospitals, but there are still two problems: First, under the multi-center screening mode, there is no Artificial intelligence joint modeling method for genetic metabolic diseases; secondly, neural network is a commonly used and effective model in the existing artificial intelligence aided diagnosis platform, and a bottleneck in extending it to multi-center joint modeling is the large number of parameters in model fusion and gradient information put pressure on the communication load

Since genetic metabolic diseases are a large class of genetic diseases, there will be many clinical screening or clinical scientific research modeling tasks in the scene, but the computing resources of the hospital are objectively limited and cannot bear the communication concurrency of a large number of joint modeling tasks, resulting in Multi-center joint modeling is less efficient

Images