In this paper, we introduce a normative modeling method using X-ray images to distinguish stages of osteonecrosis of the femoral head (ONFH). For ONFH, caused by impaired blood flow, early treatment is important. MRI is effective but expensive and les...
In this paper, we introduce a normative modeling method using X-ray images to distinguish stages of osteonecrosis of the femoral head (ONFH). For ONFH, caused by impaired blood flow, early treatment is important. MRI is effective but expensive and less accessible, while X-ray imaging is more accessible but struggles with early-stage diagnosis. The proposed method uses a U-Net model and a modified autoencoder for femoral head segmentation to learn normal femoral head features. Through the proposed model, we compare and analyze the distribution of ONFH by grade in latent space.