A novel clinical-radiomics model predicted renal lesions and deficiency in children on diffusion-weighted MRI

Background: Assessment of renal lesions and deficiency accurately remains critical in the diagnosis congenital anomalies kidneys urinary tracts (CAKUT) children. Advanced imaging such as Magnetic resonance Imaging (MRI) Diffusion weighted (DWI) allows structural functional insufficiency to be detected. Currently, radiomics machine learning models are being explored full-automated diagnostic tools. We aimed develop a integrated model predict Methods: A retrospective study 280 children with MRI/DWI were enrolled between 2018 202 at children’s hospital. total 1,037 features extracted from DWI images each participant, which divided into training set test (8:2 split). Using 5-fold cross-validated method, multiple algorithms employed when compared radiologist’s based on DWI, 99m Tc-labeled dimercaptosuccinic acid (DMSA) SPECT cortical scintigraphy or diethylenetriamine pentaacetate (DTPA) scan. Results: For detecting kidney lesions, LASSO + Random Forest algorithm outperformed other classifiers an accuracy 0.750 (95% confidence interval, 0.734–0.766) area under curve (AUC) 0.765 0.700–0.831). The performance did not show significant difference ( p > 0.05) bilateral unilateral by scanning. performed significantly better deficit than < 0.05). next built prediction for using signature clinical scintigraphy. ensemble had high-test 80.9% ± 4.2% sensitivity 91.7% 7.1% moderate calibration. Conclusion: An DWI-radiomic can utilized CAKUT.