Cervical Spine Fracture Detection via Computed Tomography scan

Abstract

The application of artificial intelligence in image processing and decision support in the medical field has received increasing attention recently in the community. In this work, we did experiments with multiple machine learning models to find the one that matches radiologists’ performance in detecting and locating fractures on the seven vertebrae of the cervical spine via a Computed Tomography scan. Among our experiments, the model, which consists of two stages using deep convolutional networks with RNN and Attention layers to classify whether a patient has a cervical spine fracture, achieved the highest performance.