Estimation of Vertebral Compression Fractures Based on Image Segmentation

We propose a process to assist physicians in diagnosing Vertebral Compression Fractures (VCFs), which identifies six positioning points for assessing whether vertebra is fractured. The proposed process combines deep learning models with image processing, utilizing Mask R-CNN to segment vertebral images and subsequently crop the results into individual vertebra images. Through post-processing steps, we obtain six positioning points of the vertebra. Finally, we compare our method with an exhaustive algorithm, while there is some discrepancy between the results, our method significantly reduces computational complexity compared to the exhaustive algorithm.