RPL Project:
The Application of Deep Learning Methods for Proximal Humerus
Fracture Feature Identification
Project Obejectives
Develop and validate a computer vision model capable of identifying fracture features
using X-ray images.
We developed our model using a large set of labeled X-ray images from patients with proximal humerus fractures. After careful review, only high-quality images were included, excluding cases that could interfere with accurate analysis, such as those with prior hardware or poor image clarity.
To analyze the data, we used an artificial intelligence approach that allows the model to learn from multiple X-rays for each patient rather than relying on a single image. This method helps capture a more complete picture of each injury. The model also learns to focus on the most important parts of each image depending on the specific clinical feature being evaluated, improving its ability to accurately identify fracture characteristics and support clinical decision-making.
Our Approach
Figure 1. Labelbox Interface
To analyze the data, we used an artificial intelligence approach that allows the model to learn from multiple X-rays for each patient rather than relying on a single image. This method helps capture a more complete picture of each injury. The model also learns to focus on the most important parts of each image depending on the specific clinical feature being evaluated, improving its ability to accurately identify fracture characteristics and support clinical decision-making.