节点文献

基于图像识别与拉曼光谱联合技术的药品核对系统构建

Construction on medication verification system based on the integration of image recognition and Raman spectroscopy

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 王敏刘海涛彭竹竹刘韶

【Author】 WANG Min;LIU Haitao;PENG Zhuzhu;LIU Shao;Dept. of Pharmacy, Xiangya Hospital, Central South University;Dept. of Nursing, Xiangya Hospital, Central South University;

【通讯作者】 刘韶;

【机构】 中南大学湘雅医院药学部中南大学湘雅医院护理部

【摘要】 目的 针对住院药房单剂量调剂中药品检测机对“同形异谱”药品存在识别盲区的问题,构建基于图像识别初筛与拉曼光谱确证的药品双重核对系统。方法 建立涵盖296种口服药品的图像和拉曼光谱数据库,基于余弦相似度算法(判定阈值0.95)进行光谱匹配;构建“图像特征初筛-拉曼光谱确证”双重核对系统,采用前后自身对照研究,在5个临床病区开展应用效果评价。结果 构建的双重核对系统对库内药品的平均识别准确率为99.2%,对“同形异谱”代表性品种可实现100%准确鉴别。试验组(图像识别+拉曼光谱确证核对)单包药品平均核对耗时较对照组(图像识别+人工肉眼实物比对)缩短31.5%~43.3%(P<0.001);护士对试验组操作便捷性、识别效率、核对置信度及心理压力缓解维度的满意度均显著优于对照组(P<0.01)。结论 “图像特征初筛-拉曼光谱确证”双重核对系统可有效突破传统机器视觉的技术瓶颈,在保障用药安全的同时能显著提升工作效率和工作人员满意度。

【Abstract】 OBJECTIVE To construct a dual-verification system integrating image pre-screening and Raman spectroscopy for inpatient pharmacy unit-dose dispensing in response to the issue of recognition blind spots for drugs of “same appearance but different spectrum” by drug inspection machines. METHODS An image feature and Raman spectroscopy database, covering 296 oral medications, were established. Spectral matching was performed using a cosine similarity algorithm(decision threshold 0.95). A dual-verification system of “image pre-screening and Raman spectroscopy confirmation” was designed, and a self-controlled before-and-after study was conducted across 5 clinical wards. RESULTS The system achieved a mean recognition accuracy of 99.2% for all medications in the database, with 100% accurate identification of representative “same-appearance but differentspectrum” drugs. The average verification time per-unit in the experimental group(image recognition+Raman spectroscopy confirmation and verification) was reduced by 31.5%-43.3%, compared with the control group(image recognition+manual visual comparison with actual objects)(P<0.001). Nurses’ satisfaction scores in the dimensions of operational convenience, identification efficiency, verification confidence, and psychological stress relief in the experimental group were all significantly superior to those of the control group(P<0.01). CONCLUSIONS The dual-verification system of “image pre-screening and Raman spectroscopy confirmation” effectively overcomes the technical limitations of conventional machine vision. It enhances work efficiency and staff satisfaction while ensuring medication safety.

【基金】 湖南省自然科学基金项目(No.2023JJ60520)
  • 【分类号】R95
  • 【下载频次】273
节点文献中: 

本文链接的文献网络图示:

本文的引文网络