Wound information identification system based on image recognition technology: a case of assisting home pressure ulcer wound care

• Main Authors: Yu-Jung Chen, 陳禹仲
• Other Authors: Shou-Chih Lo
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/3bdcr7

碩士 === 國立東華大學 === 資訊工程學系 === 107 === At home wound care, the daily care of pressure ulcers is usually the responsibility of the primary caregiver, and the nursing staff need to make regular pressure ulcer records. With the rapid development of image recognition and AI technology, many medical researches in recent years have proposed image recognition technology to solve the problem. Therefore, this paper hopes to combine image recognition and deep learning techniques to assist wound record personnel in reducing the time for patients to expose wounds when recording wound statuses, and to provide some useful information on wound to assist personnel analysis. The purpose of this paper is to design a wound information identification system(WIIS). For open and connected chronic wounds, after the wound is photographed together with the provided color patches, the wound information can be identified through the wound image, and the wound information can be provided to assist the recording. For example, the wound area, the proportion distribution of the wound color, etc., to achieve the auxiliary wound record, reducing the burden on the person. This paper presents a Wound contour positioning based on Deep Learning, WcpCNN (Wound contour positioning based on CNN) and WcpMLP (Wound contour positioning based on MLP). Using WcpMLP to detect wound contour, compared with the traditional Canny, can achieve the lowest wound area error. And according to the experiment, when the camera lens is parallel to the wound, the rotation will not cause an error. However, when the camera angle is tilted, the larger the angle, the greater the error caused by the rotation.