CONTEMPORARY CAPABILITIES OF ARTIFICIAL INTELLIGENCE TECHNOLOGIES IN THE DIAGNOSIS OF KIDNEY AND URINARY TRACT TUMORS: A SCIENTIFIC REVIEW

Abstract

The increasing incidence of urinary tract tumors exacerbates the problem of accurate noninvasive diagnosis and pre-treatment assessment of tumor potential. Traditional imaging and morphological verification methods are often limited in their ability to reliably predict tumor aggressiveness, leading to overtreatment of some patients and delayed detection of significant forms. In this regard, a systems analysis of modern artificial intelligence (AI) capabilities is of particular interest. This article summarizes the results of an exploratory analysis of modern experimental and clinical studies on the use of AI algorithms in the diagnosis of kidney and urinary tract tumors. It is shown that neural network models based on computed tomography, magnetic resonance imaging, ultrasound, endoscopic and histological images demonstrate high accuracy in automatic tumor segmentation, differentiation of benign and malignant neoplasms, and determination of histological types and grades of malignancy. Special attention is given to the potential of AI in predicting clinical outcomes, including assessing the risk of recurrence, survival, the likelihood of response to therapy, and functional outcomes after surgical treatment. It is noted that radiomics and radiogenomic signatures extracted using deep learning possess high prognostic value and reflect the molecular biological characteristics of tumors. The use of AI creates a new paradigm for noninvasive tumor phenotype assessment, enabling a transition from descriptive diagnostics to personalized prognosis even before treatment begins.