利用深度学习技术构建药品包装识别系统方法研究

Research on construction drug packaging identification system by deep learning

目的:在门诊药房发药流程中应用深度学习技术,讨论构建深度学习系统相关平台的模式与技术方案。方法:阐述药品包装识别系统工作原理,以及由训练、推理、开发管理三类平台构建深度学习系统的相关技术。结果:应用深度学习技术的药品包装识别系统现已取得初步成果,已完成对500种药品的分类识别,训练时验证准确度为96.4%。结论:深度学习平台在本系统中应用的相关技术与方法,对其他深度学习模型在医疗领域的应用有一定参考价值。

Objectives:To apply deep learning techniques in drug delivery process in outpatient pharmacy,and to discuss the models and technical solutions for building a deep learning system related platform.Methods:The working principle of drug packaging identification system and related technologies for constructing deep learning system by training,reasoning and development management were discussed.Results:The drug packaging identification system using deep learning technology has achieved initial results.The classification and identification of 500 kinds of drugs has been completed and the verification accuracy during training is 96.4%.Conclusion:The related technologies and methods applied by the deep learning platform in this system have certain reference value for the application of other deep learning models in healthcare industry.