Table 5 Osteoporosis prediction performance of the conventional single-decoder U-Net and the corresponding change in model performance (\(\Delta\)) compared to the proposed multi-decoder segmentation. Negative \(\Delta\) values indicate performance drops relative to our proposed framework. Averages are computed over three trials with different random seeds. The most significant performance drops are underlined for emphasis, while the highest performance for each metric is highlighted in bold.

From: Osteoporosis prediction from hand X-ray images using segmentation-for-classification and self-supervised learning