Classification detection method for distributed small-scale medical data set

Abstract

The invention discloses a construction method for classification detection for a distributed small-scale medical data set, and the method comprises the steps that: a teacher network is set in a central server, student networks are set in local servers of a plurality of medical institutions, the central server can access the medical data of the plurality of medical institutions, and the local server can only access the medical data of the corresponding medical institution; the central server carries out training on the teach network according to the medical data obtained from each medical institution and a preset task demand to generate a soft label; the local server carries out training on a student network according to the medical data of each medical institution to generate a hard tag, determines a real tag corresponding to the medical data, and calculates a network loss function according to the soft tag, the hard tag and the real tag; and network loss is calculated according to the loss function, and the teacher network and the student network are optimized according to the network loss to obtain the distributed knowledge distillation network. The technical problem that in the prior art, a trained learning model is poor in performance is solved.

Description

technical field [0001] The present application relates to the technical field of medical data processing, in particular to a classification and detection method for distributed small-scale medical data sets. Background technique [0002] Privacy issues are currently involved in many fields, especially in the medical field. Many countries and medical institutions have implemented relevant legal protection and review mechanisms to prevent malicious copying or even tampering of sensitive data of medical patients. However, these regulations are like a double-edged sword. Although they can protect the privacy of users, they will also objectively cause insufficient collaboration and data sharing between health records. [0003] Furthermore, with the rapid development of machine learning technology, in order to analyze medical data better, more conveniently and quickly, it is necessary to train a successful machine model, and training a machine model requires a sufficient amount of...

AI Technical Summary

Problems solved by technology

[0004] The technical problem solved by this application is: Aiming at the poor performance of the learning model trained in the prior art, this application provides a distributed small-scale medical The construction method of the classification detection of the data set, in the scheme provided by the embodiment of this application, proposes a distributed knowledge distillation network, in the case that the original medical data is not shared between the student networks, that is, while ensuring the privacy of the medical data Under certain circumstances, the medical data distributed in different student networks is used to learn and guide training through the teacher network, so as to avoid a large amount of valuable data in various medical institutions and become a "data island", and all the data distributed in different places can be learned The characteristics and distribution of the data enrich the amount of training data, and the performance of the model is higher than that of the model trained only from a single small amount of data.

Images