Using hyperspectral imaging to predict the occurrence of delayed graft function

Delayed Graft Function (DGF) is a prevalent complication in kidney transplantation (KT) that significantly affects allograft function and patient prognosis. Early and precise identification of DGF is crucial for improving post-transplant outcomes. In this study, we present KGnet, a predictive model leveraging hyperspectral imaging (HSI) to assess delayed graft function status. We analyzed 72 zero-hour biopsy samples from transplanted kidneys with confirmed pathological diagnoses, capturing spectral data across a wavelength range of 400 to 1000 nm. By examining spectral signatures related to tissue oxygenation, perfusion, and metabolic states, our approach enabled the detection of subtle biochemical changes indicative of DGF risk. The preprocessed spectral data were input into KGnet, achieving a prediction accuracy of 94 % for DGF occurrence, significantly outperforming existing predictive models. This study identifies key spectral signatures associated with DGF, allowing for precise risk prediction even before clinical symptoms emerge. Leveraging HSI for early detection introduces a novel pathway for individualized post-transplant management, offering substantial potential to enhance kidney transplantation outcomes and patient quality of life. These findings highlight significant clinical and research implications for the broader application of HSI in transplant medicine.