Table 1 A summarized overview of existing techniques related to spine analysis.
From: Transformer based spinal vertebrae localization and scoliosis curvature classification
Paper Title | Key Contribution | Focus | Dataset | Gaps | Results |
|---|---|---|---|---|---|
Method tailored for automatic Cobb angle detection using deep learning techniques | Cobb angle detection | AASCE MICCAI 2019 | Lack a comprehensive comparison with existing Cobb angle detection methods | SMAPE score of 25.69 | |
Method tailored for accurately detecting vertebra landmarks in scoliosis assessment | Vertebral Landmark Detection | AASCE MICCAI 2019 | Network skips vertebrae with lower morphology properties than others | SMAPE score of 10.81 | |
Novel approach for automated estimation of spinal curvature | Spine Curvature Estimation | AASCE MICCAI 2019 | Method has a long running time due to smoothing with Euler method | SMAPE score of 22.96 | |
Novel approach for automated spinal curvature estimation | Spine Curvature Estimation | AASCE MICCAI 2019 | Limited impact of different network architectures or loss functions | SMAPE score of 21.71 | |
Consistency learning approach for joint spine segmentation and Cobb angle regression | Spine Segmentation and Cobb Angle Estimation | AASCE MICCAI 2019 | Lack of exploration of the impact of different consistency learning strategies on model performance | SMAPE score of 7.32 | |
Multi-task learning method for directly estimating spinal curvature, reducing reliance on intermediate steps | Spine Curvature Estimation | AASCE MICCAI 2019 | Limited exploration of the method’s performance on diverse datasets | SMAPE score of 12.97 | |
Novel consistency loss function tailored for more accurate Cobb angle estimation | Cobb angle detection | AASCE MICCAI 2019 | Limited exploration of the method’s performance on diverse datasets | SMAPE score of 8.62 | |
Novel Linformer-based approach for Cobb angle rectification, enhancing the efficiency of angle correction | Cobb angle detection | AASCE MICCAI 2019 | Limited exploration of the method’s performance on diverse datasets | SMAPE score of 7.91 |
