Timing sequence deep survival analysis system with active learning

Abstract

The invention discloses a time sequence deep survival analysis system with active learning. The system comprises a data acquisition module, an active learning module and a time sequence deep survivalanalysis module. The data acquisition module is used for acquiring survival data of a to-be-analyzed object; the active learning module selects part of right deletion data to mark survival time in combination with an active learning method; and the time sequence deep survival analysis module constructs a time sequence deep survival analysis neural network model, and takes the undeleted data and the right deleted data as model inputs to obtain a survival time prediction result of the to-be-analyzed object. According to the method, the right deletion data and the time sequence characteristics inthe survival data can be fully utilized. Compared with a traditional survival analysis model, the system has the advantages that the problem that high-dimensional data is difficult to process is solved, and the problem that the model is poor in performance under the condition that only a small amount of data is not deleted in survival analysis is solved; meanwhile, extraction and utilization of data time dimension features are increased, the application range of the model is expanded, and the expression effect of the model is improved.

Description

technical field [0001] The invention belongs to the medical field, and in particular relates to a time series deep survival analysis system combined with active learning. Background technique [0002] Survival analysis refers to an analysis method that comprehensively analyzes whether an event will occur and when it occurs. It can study the relationship between survival time, outcome and many influencing factors and their degree. It is widely used in medicine, biology, finance and other fields. It is widely used and has very important research significance and research value. The goal of survival analysis is to analyze and predict the occurrence of events on the observed objects through the observation of events that have already occurred. Through survival analysis, we hope to obtain the characteristics of the survival of the research object in terms of time, estimate the survival rate and average survival time, and then study the protective and unfavorable factors that aff...

AI Technical Summary

Problems solved by technology

[0007] However, due to various limitations in the means of observation or the observed objects, the electronic medical record data has a clear time point of the outcome event, that is, the number of observed patient deaths is small, and a large part of the patients are still alive due to discharge, transfer, loss of follow-up or so far. Right-censored survival data due to some reasons, and the right-censored data may account for more than 90% of the survival data

The existing survival neural network represented by Deep Surv does not perform well when dealing with survival analysis problems based on electronic medical records with a high proportion of censoring, which greatly limits the application of the model in the field of medical and health

[0008] On the other hand, medical electronic health data is often time-series, and the characteristics are closely related to time. However, neither the traditional Cox proportional hazards model nor the survival neural network based on the convolutional neural network represented by Deep Surv can make use of the data. The timing characteristics of

Images