International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 06 | Jun 2024
www.irjet.net
p-ISSN: 2395-0072
Detection and Classification of Femoral Neck Fracture using YOLOv8 Mousab Abibi Abdi1, Rafet Akdeniz2, 1MS Student, Computer Engineering, Istanbul Aydin University, Istanbul, Turkey 2Professor, Computer Engineering, Istanbul Aydin University, Istanbul, Turkey
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - : Femoral neck fractures represent a critical
revolutionize the approach by allowing for more uniform and algorithm-driven analysis. These advances in AI-driven approaches offer great ways for better-automated image segmentation and local feature extraction, which may replace subjective radiographic union scoring with a standard and portable algorithm-based evaluation of severe injuries like femur fractures.
orthopedic problem to their complexity and possible complications. For this reason, it is crucial to detect these fractures quickly and accurately for proper clinical management. This study use YOLOv8 model to detect and classify femoral neck fractures in X-ray images. The performance of YOLOv8 is 97.9% in mAP50, 93.5 in precision and 62.5% in mAP50-95. Our proposed system consist data collection, preprocessing, training and testing the model and model deployment. The proposed model shows potential for automated detection and classification of femoral neck fractures which provides valuable assistance to radiologists.
Studies have found that the use of deep learning techniques in femoral neck fracture detection is promising. In recent research, advanced deep-learning techniques were implemented to automate the detection and classification process for femoral neck fractures, with an accuracy rate of 92.3% in two-class prediction cases and 86.% in three-class prediction [5]. In another study, deep learning and genetic algorithm approaches were used, in which a sensitivity value of 83% was recorded, a specificity value of 73%, and an F1 score value of.78, respectively [6]. The level of accuracy achieved in this particular study shows that combining genetic algorithms with deep learning can enhance fracture detection.
Key Words: Femoral neck fracture, deep learning, YOLO, YOLOv8, X-Ray images.
1.INTRODUCTION Femoral neck fractures are a significant cause of death and disability in the elderly [1], [2]. While examining pelvis radiographs, slipping femur fractures are sometimes missed, and their delayed diagnosis means an increase in costs as well as harmful outcomes [3]. Consequently, the effectiveness of any recommendation for treatment may depend on a prompt and accurate diagnosis. In practice, doctors and radiologists use X-ray images to find fractures. Detecting these fractures through manual checks or with the help of a conventional X-ray machine is a laborious and timeconsuming process.
Furthermore, the YOLOv8 algorithm is applied for the detection of fractures in pediatric wrist trauma x-ray images, attaining a mean average precision (mAP) of.638% under an overlap of 50% [7]. This is an indication that YOLOv8 performs well in identifying fractures within pediatric cases accurately. Another deployed YOLOv8 model on femur fracture detection gave out an mAP score of 0.842 at a 50% overlap threshold for both precision 0.85 and recall 0.83 [8].
In four non-Level-1 trauma hospital emergency rooms, 975 radiography patients with subsequent CT1 showed 68 falsenegative cases. The greater trochanter, ilium, and pubis are the areas with the most often overlooked fractures [3]. Osteoporotic hip fractures account for 14% of those occurring in the United States of America (USA), but they represent 72% of costs related to healthcare caused by osteoporotic fractures [3]. Costs are the other part of the story. Patients who cannot walk by themselves are 40% after a hip fracture; the other 60% cannot do at least one daily living activity, while 80% have additional limitations such as no driving anymore after these kinds of fractures have affected their pelvises.
2. PROPOSED SYSTEM
Fig-1: Proposed System Our proposed femoral neck fracture detection system using YOLOv8 is based on a well-structured machine learning pipeline. The first step involves collecting data by acquiring original radiographs of femoral necks. This is then followed by preprocessing, where images undergo various enhancement procedures, such as image augmentation. In
Additionally, it is widely accepted that current technologies such as x-ray, CT scan, or MRI are not always enough to diagnose severe healing abnormalities [4] like delayed unions or non-unions in patients with femur fractures. There is a necessity to introduce AI-based solutions to
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 1210