基于视频目标检测的中药饮片调剂预警系统  

Early warning system for dispensing Chinese herbal slices based on video object detection

作　　者：王艳 唐燕 石琳 张林 高玉培 王浩霖 杨丰蔚 韩爱庆 WANG Yan;TANG Yan;SHI Lin;ZHANG Lin;GAO Yupei;WANG Haolin;YANG Fengwei;HAN Aiqing(Beijing University of Chinese Medicine,Beijing 102488)

机构地区：[1]北京中医药大学管理学院,北京102488

出　　处：《北京生物医学工程》2024年第1期43-50,共8页Beijing Biomedical Engineering

基　　金：教育部产学合作协同育人项目(201802003031)资助。

摘　　要：目的构建基于视频目标检测的中药饮片调剂预警系统,降低中药饮片调剂混淆事件发生率,保障用药安全。方法设计基于人工智能的中药饮片调剂预警系统开发方案,分为视频采集、目标检测、文字识别和预警四个模块。预警系统可以驱动监控视频实时检测药剂师中药饮片调剂内容,与处方实时比对,对漏配、错配和多配等错误调剂信息进行语音和图文信息预警。本文以小柴胡汤加减方为基础,采购3个批次中药饮片样本,采集并标注1524张饮片图像,制作中药饮片数据集,基于Faster R-CNN算法训练目标检测模型并进行系统构建。结果经测试Faster R-CNN模型均值平均精度(mean average precision,mAP)达到了9510%,利用训练好的目标检测模型结合文字识别和预警算法构建预警系统,对中药饮片调剂过程进行识别,系统能对调剂错误行为进行准确自动报警。结论该系统可实时、主动进行中药饮片调剂检测和预警,为中药饮片调剂智能化提供新思路,提高中药饮片调剂科技水平,促进中医药与人工智能相结合。Objective To construct an early warning system for the dispensing of Chinese herbal slices based on video object detection,and to reduce the incidence of confusion events and ensure the safety of medication.Method Designed an artificial intelligence of Chinese herbal slices dispensing early warning system development scheme,divided into video collection,object detection,optical character recognition and early warning of four modules,early warning system could drive the surveillance video real⁃time detection of pharmacist Chinese herbal slices dispensing content,real⁃time comparison with the prescription,wrong dispensing information for voice and graphic information early warning.This article was based on the addition and subtraction of xiaochaihu decoction,3 batches of Chinese medicine tablet samples were purchased,1524 herbal pieces images were photographed and annotated,and a dataset of Chinese herbal slices was produced.Results After testing,the mean average precision(mAP)of the Faster R⁃CNN model achieved 9510%.The trained object detection model combined optical character recognition and early warning algorithm to build an early warning system to identify the process of dispensing Chinese herbal slices,and the system could accurately and automatically alarm the wrong behavior of the dispensing.Conclusions The system can carry out real⁃time and active detection and early warning of the adjustment of Chinese herbal slices,provide new ideas for the intelligent adjustment of Chinese herbal slices,improve the scientific and technological level of Chinese herbal slices,and promote the combination of traditional Chinese medicine and artificial intelligence.

关 键 词：目标检测 中药饮片 调剂 差错 预警 

分 类 号：R318[医药卫生—生物医学工程]