( Kidney International October 2021, Volume 100Issue 4SupplementS1-S276, KDIGO 2021 Clinical Practice Guideline for the Management of Glomerular Diseases)
A multicenter study to develop a non-invasive radiomics model to identify UTI stone in vivo using machine learning.
Like other omics technologies, radiomics uses high-density data to characterize a disease or disease process. Specifically, in the radiological domain, quantitative data is extracted from radiographic images to identify features of disease that are not easily visualized. Zheng et al. applied radiomics techniques followed by machine learning on kidney stones to differentiate infection from non-in vivo stones. CT scans were analyzed and 1,300 features were extracted from each stone image. More than 1,000 patients with stones constituted training and validation cohorts and stone composition was verified by spectroscopy after surgical removal. Ultimately, the researchers coupled their radiomic signature to urine pH and the presence or absence of urease-producing bacteria in the urine to develop a model that exhibited excellent discrimination between urine stones. infection and non-infection. This non-invasive procedure can improve stone management.