Bone Fracture Detection Using Deep Learning

Abstract

According to statistics, more than 65% of the population in Delhi is at high risk of bone fractures or is already diagnosed with one. (Study conducted during 2018). Broken bones or bone fractures occur when an overwhelming amount of force is applied to a bone, which is way stronger than what a bone can withstand. This troubles the strength and the structure of the bone which will lead to an immense amount of pain and a lot of treatment to get well. When bone fractures are untreated, it will lead to numerous complications such as a nonunion or delayed union. In some cases, the bone doesn’t heal at all, which means that it will remain broken until it has been treated. As a result of this, the swelling and the pain will continue to worsen over time. Most fractures heal in 6-8 weeks, but this varies tremendously from bone to bone and in each person based on many of the factors discussed above. Hand and wrist fractures often heal in 4-6 weeks whereas a tibia fracture may take 20 weeks or more. Doctors can usually recognize most fractures by examining the injury and taking X-rays. But in some cases, it is hard for them to diagnose it. This study focuses on automating the process of detecting fractures from X-rays through deep learning. The model has achieved an accuracy of almost 95%. Though that is true in the paper, The model won’t perform that accurately when it comes to real-world applications as it depends on various factors such as X-ray shot style, the angle at which it has been shot etc. This model can be fine-tuned to serve real-world applications and the need for these kinds of deep learning models in our society is undoubtedly high as the society is collectively progressing towards a better future aided by artificial intelligence, deep learning and machine learning models..