VinDr-SpineXR allows classifying between normal and abnormal spine X-ray images with 90% accuracy on the AUC measure (area under the curve indicates the ability to distinguish between patients with disease and not sick). In addition, the model is also able to accurately localize 06 types of common spinal injuries in Vietnam, including: bone spurs; disc stenosis; surgical materials; narrowing of the adapter hole; spondylolisthesis and vertebral collapse with mAP = 0.55 (mAP is the average accuracy index, which measures the performance of object detection models in images).
VinDr-SpineXR: A deep learning framework for spinal lesions detection and classification from radiographs