This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3029039, IEEE Access VOLUME XX, 2020 Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000. Digital Object Identifier 10.1109/ACCESS.2017.Doi Number Ground Truth Annotated Femoral X-ray Image Dataset and Object Detection Based Method for Fracture Types Classification Yubin Qi 1 # , Jing Zhao 2 # , Yongan Shi 3 , Guilai Zuo 1 , Haonan Zhang 1 , Yuntao Long 1 , Fan Wang 1 , Wen Wang 1 * 1 Department of Orthopedics, The First Affiliated Hospital of Shandong First Medical University, Jinan, Shandong 250014, China 2 Shanghai Key Laboratory of Multidimensional Information Processing, East China Normal University, Shanghai 200241, China 3 Department of Orthopedics, Shandong Wendeng Osteopathic Hospital, Weihai, Shandong 264400, China #Equal contribution *Corresponding author: Wen Wang (e-mail: wangwenqfs@sina.com). This work is supported by the Science and Technology Commission of Shanghai Municipality (No.19511120800). ABSTRACT Precise classification of femoral fractures contributes to accurate surgical strategies and better prognosis after surgery. An effective and accurate system for diagnosing femoral fractures and classifying its types will play a vital role in clinical work. This work aims to achieve the automatic detection and classification of femoral fractures in X-ray images. We build a benchmark which includes 2333 X-ray images with 9 different fracture types, and each of them is manually labeled with the ground truth boxes indicating the femoral shaft fractures and its corresponding categories according to the Association for the Study of Internal Fixation (AO). An anchor-based Faster RCNN detection model, with the backbone of ResNet-50 being constructed in a multi-resolution feature pyramid networks (FPN), is used for locating fractures regions and classifying its types. The total image level accuracy reaches 71.5%, which is higher than some of the orthopedic surgeons can get, especially young orthopedic surgeons. Therefore, it is practicable to take advantage of artificial intelligence to detect and classify the femoral shaft fractures. INDEX TERMS orthopedic procedures; orthotics; image classification; image processing; image analysis I. INTRODUCTION In orthopedic high-energy injuries, femoral fractures are very common, and its incidence has been reported as an average of up to 37 per 100000 patient years [1]. Different types of fractures can be found in femoral shaft fractures according to the location of the fracture, the degree of fracture crushing, the degree of muscle and soft tissue damage, and the injury mechanism. The current gold standard for the treatment of femoral shaft fracture is intramedullary nailing, introduced by Groves in the United Kingdom and by Kuntcher in Germany [2]. However, with the wide application of intramedullary nail fixation, there are many complications, such as delayed union, nonunion, fracture of intramedullary nail and re-fracture. The reason is largely related to the classification of fracture and the surgical technique of doctors. Different treatment options are selected in accordance with different fracture types. Reasonable and accurate fracture classification is very important for the choice of surgical schemes. Therefore, an 1 effective and accurate fracture diagnosis and classification plays a vital role in clinical work. Fracture classification systems are now mostly dependent on three-dimensional computed tomography (CT), X-ray and magnetic resonance images (MRI) and manual visual inspection according to AO classification [3]. However, the information of these images maybe accurate enough for surgeon to detect some clear fraction but not to examine small fractures because of their low resolutions [4]. Therefore, there are several computer-aided algorithms involved to assist in classifying the fractures and supply more quantitative information. Some researchers present 3D interactive software system to help surgeons watch the organ more clearly and improve the clinical knowledge in orthopedics [5]. In addition, there are more research works