Identity-Consistent Diffusion Network for Grading Knee Osteoarthritis Progression in Radiographic Imaging

TL;DR

The paper tackles KOA prognosis by producing both a predicted future knee X-ray and a progression severity grade, addressing the lack of visual explainability in prior methods. It introduces IC-RDN, a diffusion-based generator guided by an identity prior and a downstream progression predictor, to generate a $12$-month X-ray from baseline images and to forecast KOA progression using both images. The identity prior is learned via contrastive triplet loss and integrated into diffusion through cross-attention, while the progression predictor leverages a two-stream CNN and an MLP classifier. Evaluated on the Osteoarthritis Initiative (OAI) data, IC-RDN delivers competitive visual generation quality and improved progression-grade predictions, highlighting the practical potential of combining generated radiographs with baseline images for multifaceted KOA prognosis.

Abstract

Knee osteoarthritis (KOA), a common form of arthritis that causes physical disability, has become increasingly prevalent in society. Employing computer-aided techniques to automatically assess the severity and progression of KOA can greatly benefit KOA treatment and disease management. Particularly, the advancement of X-ray technology in KOA demonstrates its potential for this purpose. Yet, existing X-ray prognosis research generally yields a singular progression severity grade, overlooking the potential visual changes for understanding and explaining the progression outcome. Therefore, in this study, a novel generative model is proposed, namely Identity-Consistent Radiographic Diffusion Network (IC-RDN), for multifaceted KOA prognosis encompassing a predicted future knee X-ray scan conditioned on the baseline scan. Specifically, an identity prior module for the diffusion and a downstream generation-guided progression prediction module are introduced. Compared to conventional image-to-image generative models, identity priors regularize and guide the diffusion to focus more on the clinical nuances of the prognosis based on a contrastive learning strategy. The progression prediction module utilizes both forecasted and baseline knee scans, and a more comprehensive formulation of KOA severity progression grading is expected. Extensive experiments on a widely used public dataset, OAI, demonstrate the effectiveness of the proposed method.

Figures (6)

• Figure 1: Pipeline for KOA severity progression grading, encompassing a) predicted future knee joint X-ray scan, and b) predicted KOA progression severity grade.
• Figure 2: Illustration of the proposed IC-RDN method. (a) An identity prior module formulates identity representation, which guides the diffusion process for the generation. (b) An I2I image generative network to forecast a future 12-month X-ray scan. (c) A downstream KOA progression prediction network, taking the baseline and the predicted 12-month X-ray scans as inputs.
• Figure 3: Illustration of t-SNE analysis between IC-RDN (ours) and I2I diffusion model.
• Figure 4: Comparison of generated X-ray images from Cycle GAN zhu2017unpaired and our proposed IC-RDN on knee joints i-viii.
• Figure 5: Generative results from IC-RDN among different patients' joints from i to xxii. The detailed joint structure comparison between baseline visit images, ground truth and results from IC-RDN shows our model's ability to generate progression patterns.