以影像識別為基礎之藥錠比對方法與裝置之研發

Development of an Intelligent Drug Identification Method and Device Based on Digital Image Recognition

近幾年，社會出現護理人力嚴重不足的問題，而護理人力不足也導致病人受護品質降低與死亡風險的提高；以給藥為例，“給藥錯誤”事件在世界各國的醫療不良事件中皆高居首位，一旦遇到藥發錯，護理人員很容易造成醫療糾紛而身陷囹圄。過去僅有少數文獻進行藥物影像辨識研究，未見有針對此應用的研究被提出；為此，本研究進行一項以藥錠核對為目的之藥錠影像比對方法與裝置的研發，期望透過此裝置協助護理人員於給藥前進行藥錠核對，達到降低人為疏失，提高病人用藥安全的目的。為了要做到藥錠辨識自動化功能，本研究以電腦視覺技術為基礎，設計所需的裝置與影像處理演算法。方法上則使用藥錠形狀、顏色與壓印等影像特徵資訊來做為辨識依據；其中，演算法包含使用簽名特徵來描述形狀、色彩量化估算藥錠顏色、光學字元辨識來識別壓印與自行設計之影像處理方法，本研究所使用的方法，僅以形狀或顏色來進行辨識，準確度高達100%，壓印文字辨識率的部分則維持在88%左右，在文章實驗的部分展現本研究所開發的演算法與裝置的可行性。

In the recent years, there has been an impact on the quality of patients' care due to human negligence, especially in care institution. Nursing human resourcesare directly related to patient safety. For example, nurses need to be individually checked drugs and distributing drugs to multiple patients. For medical purposes, most hospitals introduce a large number of drugs, which nurses cannot remember completely. In some units of hospital, such as intensive care unit, drug identification is performed just by naked-eye examination. It is difficult for even a well-trained and experienced nurse to perfectly identify various drugs without the assistant of equipments. Many literatures examining the application of technology in the medical industry conclude that the quality of medical care would be enhanced significantly through the application of technology like automated machines. Therefore, the intelligent drug identification algorithms and a device based on digital image recognition will be developed to assist nurses in drug identification in this study. The device is based on digital image processing. The algorithms include color feature extraction (color quantization), shape feature extraction (signature), and imprint feature extraction (optical character recognition). Using the proposed automatic drug identification device, nurses can enhance the simplicity and the safety of nursing care in addition to increasing convenience and reducing human error.