SSL-CL-RUS - A Semi-Supervised Framework for Renal Ultrasound Segmentation in CKD: Combining Pseudo-Label Guided Contrastive Learning with ST+.

Chronic kidney disease (CKD) is a pervasive global health challenge that necessitates early and accurate diagnosis for effective management. Renal ultrasound imaging is integral to CKD evaluation; however, precise segmentation of kidney structures is often compromised in advanced stages due to degraded image quality. In this work, we propose a novel semi-supervised segmentation framework that synergistically integrates Pseudo-Label Guided Contrastive Loss (PLGCL) with the ST++ learning strategy. By leveraging both labeled and unlabeled data, our method enhances segmentation accuracy. Extensive experimental evaluations demonstrate that our approach outperforms state-of-the-art techniques, achieving significantly higher Dice coefficients across all CKD stages, with marked improvements observed in late-stage disease (G4 and G5). These findings underscore the potential of combining contrastive learning and pseudo-labeling to address the challenges of ultrasound image segmentation in complex clinical scenarios.Clinical Relevance- The proposed framework enhances the delineation of renal structures in ultrasound imaging, particularly in advanced CKD stages, thereby supporting more accurate assessments of disease progression. Improved segmentation fidelity can facilitate optimal treatment planning and clinical decision-making, ultimately contributing to better patient outcomes through more reliable morphological evaluations of the kidneys.