Medical term automatic standardization system and method integrating self-supervision and active learning

Abstract

The invention discloses a medical term automatic standardization system and method fusing self-supervision and active learning. The system comprises a candidate set generation module, a self-supervised learning module for training a term standardization model, an active learning module, a precise sorting module for comprehensively evaluating a term standardization model prediction result from text and semantic dimensions, and other basic modules. The system further comprises optimization modules such as a semi-supervised learning module and a direct superior term retrieval module. According to the method, an automatic medical term standardized model can be realized under the condition of less annotation data, the model keeps the capability of rapid updating and upgrading, and the workload of manual intervention is greatly reduced while the accuracy of an output result is ensured; the new clinical concept can be matched with the immediate superior terms, and an accurate position can be found in the standard term table, so that the completeness and uniformity of a standardization result are ensured.

Description

Technical field [0001] The present invention belongs to the technical field of Chinese medical terms and multi-center medical information platform technology, and in particular, the present invention relates to a medical term automatic standardization system and method for fusion self-supervision and active learning. Background technique [0002] With the popularity of electronic medical records, large amounts of medical related important information stored in various medical information systems in electronic forms, these data are clinical auxiliary diagnosis, drug development, public health monitoring and evaluation, infectious disease epidemic warfare, personality Create a huge value. Medical data standardization is a key step in promoting the integration of domestic medical systems and achieves medical data synergies, large-scale analysis. Standardization of medical terms is the first problem that the medical data standardization process is first solved. Internationally has co...

AI Technical Summary

Problems solved by technology

The terminology standardization schemes in the past were generally based on the following two ideas: (1) Use manual methods and invite professional clinicians to map and proofread surgical terms one by one, but the order of magnitude of surgical terms contained in each medical information system is in the tens of thousands, The working hours of clinicians for proofreading are very long, and it is difficult to quickly promote it in China, which hinders the rapid implementation of domestic medical data standardization; in addition, due to the differences in work experience among doctors, there is a lack of uniform standards for the mapping of surgical standard terms. It is difficult to ensure the unification of standards among different doctors, and at the same time, there are manual errors in the mapping results, and it is difficult to guarantee the unification of the mapping standards at different times for the same doctor

(2) Training medical concept semantic matching models based on machine learning algorithms, but manual labeling of data is difficult and time-consuming, resulting in insufficient training data, and the final output model has low generalization ability. In order to ensure the actual use of terminology standards The accuracy of the results requires a lot of manual verification of the output results

When the concepts generated by actual clinical operations cannot find equivalent terms with the same meaning in the standard terminology set, it is necessary to accurately locate its immediate superior standard terms, and the existing methods have failed to solve this problem well, resulting in new clinical concepts that cannot be integrated into a common standard terminology

Images