Paper 13407-105
Paper 13407-105
Explainable deep learning for rib fracture detection in chest x-rays
19 February 2025 • 5:30 PM - 7:00 PM PST | Golden State Ballroom
Abstract
This study investigates improving the accuracy of rib fracture detection using modern deep learning models and highlights the thought process behind them. Despite advancements in existing models, they often lack reliability and explainability. By implementing GRAD-CAM for visualizing regions of interest and utilizing the DenseNet121 architecture, we achieved a comparable model that is both accurate and explainable. Through data preprocessing and extensive hyperparameter tuning, our model achieved 75% training accuracy and 74% testing accuracy on the CheXpert dataset with an AUROC score of 0.82 and then an 85% validation accuracy on the Medical College of Wisconsin (MCW) dataset. The study highlights the need for further improvements in model focus and localization abilities to ensure reliable automated rib fracture detection.
Presenter
Conner Rutherford
Milwaukee School of Engineering (United States)
Connor Rutherford is currently pursuing a Bachelor of Science degree in Computer Science at the Milwaukee School of Engineering. He is a former software engineering intern at Uline and will transition to a full-time Software Developer Associate role at Uline upon graduation. In his current position, he applies his technical expertise in a real-world setting, contributing to a team of full-time software developers. His responsibilities include participating in scrum methodology, collaborating on software development projects, and contributing to the design, testing, and optimization of applications. Connor’s work emphasizes practical problem-solving and continuous learning in a fast-paced environment. Through his education and internship, Connor is building a solid foundation in computer science, demonstrating his commitment and enthusiasm to becoming a leading expert in AI and software development.