A two-level system for differential diagnosis solid renal tumors based on multiparametric ultrasonic and tomographic analysis: development and internal validation of criteria
Abstract
Differential diagnosis of solid renal tumors at the stage of standard radiological search is associated with a high frequency of overlapping imaging features, forming “gray zones” between benign and malignant histotypes. Objective: to develop and validate a two-tiered differential diagnosis system for solid renal tumors based on the integration of literature data and analysis of our own clinical data (395 patients). Materials and Methods. Based on a systematic analysis of metadata from 42 specialized publications (12,000 patients), the primary criteria of the model were formulated. Clinical testing of the algorithm was performed on our own sample of 395 patients, including clear cell carcinoma (ccRCC; n = 258), angiomyolipomas (AML; n = 35), papillary carcinoma (pRCC; n = 32), oncocytomas (n = 30), chromophobe carcinoma (chRCC; n = 28) and a control group of Bosniak IIF/III dense cysts (n = 12). All patients underwent ultrasound with color Doppler imaging, multiphase MSCT and multiparametric MRI (mpMRI). Mathematical analysis was performed using binary logistic regression and ROC analysis with bootstrap validation (B = 1000). Results. A two-component code matrix for the calculator was developed, allowing for discrete separation of diagnostic stages over time. The first level (ultrasound + MSCT) automatically segregates up to 85% of classic patterns. If signs of a “gray zone” are detected, the session is archived under a patient ID until the second level (mpMRI) is performed. Differentiation of atypical tumors at the expert stage is achieved through the integration of functional MRI markers and the calculation of critical cutoff points for the measured diffusion coefficient (MDC) for polar histotypes. The overall diagnostic accuracy of the final two-level model on the validation set was 91.3%. Conclusion. The two-level discrete approach minimizes diagnostic errors in “gray areas” of imaging, optimizes patient routing, and reduces the incidence of unnecessary surgical interventions.