Detection And Recognition Of Orthopedic Surgical Instruments Based On Deep Learning

In recent years,with the increase in the number of operations in hospitals and the improvement of health industry standards,the task of supplying and processing medical equipment required by various departments in hospitals has become more and more arduous.Due to the wide variety of surgical instruments,the inventory and inspection process is cumbersome and labor-intensive,and manual processing is not only time-consuming and labor-intensive,but also easily increases the risk of infection and even leads to the possibility of sorting errors.Under the above problems,using orthopedic surgical instruments as the research object,based on the YOLO V3 network,research is carried out to build an orthopedic surgical instrument identification and detection platform,which can realize the detection,identification,and experimental testing of orthopedic surgical instruments under the guidance of visual information and assist medical staff to better complete inventory operations.The paper will introduce the following aspects:For model training,a special data set for the detection of orthopedic surgical instruments provided by the hospital has been established.To realize real-time detection of orthopedic surgical instruments,a detection platform for orthopedic surgical instruments has been designed and built,and a detection software interface for orthopedic surgical instruments based on the Py Qt5 framework has been built into its application scenario.A thick and thin double-classification model was designed and developed for a variety of similar instruments in the orthopaedic surgical instrument package.The rough classification uses the YOLO V3 network model to quickly identify and locate the orthopedic surgical instruments in the orthopedic surgical instrument data set,and the sub-classification uses a threshold segmentation-based sub-classifier to quickly subdivide the areas of orthopedic surgical instruments that are difficult to distinguish.In order to improve the recognition and detection effect of orthopedic surgical instruments,a target detection model with better performance is proposed based on the rough classification of the model.The algorithm uses Res2Net50 as the basic backbone network and uses the SPP module to fuse multi-scale information;in the feature utilization part,the MIEB module and the LRM module are proposed to improve the model performance;the K-means algorithm is used to recalculate the a priori frame size.According to the experimental results,the effectiveness of IYOLO on the orthopedic surgical instrument dataset and the public dataset VOC2007 is illustrated.