#6073 MAGNETIC RESONANCE RADIOMIC ANALYSIS FOR THE PREDICTION OF GRAFT FAILURE IN PATIENTS WITH KIDNEY TRANSPLANTATION

Abstract Background and Aims Prediction of future graft failure with non-invasive investigations is a relevant objective for transplant clinicians. We previously demonstrated promising discrimination capacity interstitial fibrosis / tubular atrophy (IFTA) >50% in kidney biopsy machine learning (ML) based magnetic resonance imaging (MRI) radiomic algorithm. Aim the present study to evaluate accuracy MRI radiomic-based ML algorithms predicting failure. Method Single-center retrospective observational cohort on characterized recipients who underwent within 6 months from biopsy, both clinical indication, at “Azienda Ospedaliero-Universitaria di Modena”, Italy, 1/1/2012 1/3/2021. The primary outcome was identify best combination features prediction during follow-up. Segmentation renal parenchyma, cortex medulla sequences performed using 3DSlicer software. Radiomic were then extracted an in-house software pyradiomics applying Wavelet Gaussian filters. LASSO algorithm employed select correlated summarize them signature. Using signature alone merged meaningful data we trained ML-algorithms 70% cases training/validation, 10-fold internal cross-validation, 30% model testing. Model performance assessed AUC 95% confidence interval (CI). Results Seventy coupled tests (63 patients) included randomly subdivided into training/validation (n = 50) test 20). Median follow-up 24.73 (interquartile range 13.64-46.57). had 0.88 (95%_CI 0.70-0.97) training 0.57 0.23-0.86) cohort. Radiomic-clinical mixed 0.90 0.73-0.98) 0.66 0.33-0.88) Conclusion produced good ability patients transplant. Given limited number enrolled patients, validation larger (and ideally prospective) cohorts required confirm our findings. Comparison histological parameters (i.e., IFTA) currently under assessment.