Tortuosity of the coronary, cerebral arteries, aorta and its branches remains an important vascular problem, which, on the one hand, complicates selection of the X-ray surgical treatment tactics, and on the other hand worsens the disease outcome. The lack of common standards for assessment of tortuosity of the coronary, cerebral arteries, aorta and its branches reduces the diagnosis accuracy in patients at high risk of cardiovascular events. The use of machine learning for automated tortuosity assessment represents one possible solution to this problem. The study aimed to analyze and compare accuracy, feasibility, and limitations of the available methods for automated assessment of tortuosity of the coronary, cerebral arteries, aorta and its branches using the machine learning tools. The systematic review was conducted in accordance with the PRISMA protocol. The search for papers published in 2015–2025 in the PubMed, Scopus, and eLibrary databases was performed using the following keywords: deep learning, machine learning, artificial intelligence, vessel tortuosity, curvature. Six papers out of 240 were included in the analysis. The analysis has shown that 80% of approaches are based on convolutional neural networks, and skeletonization aimed to isolate small blood vessels from the artery represents an essential preprocessing phase. In 50% of papers, tortuosity was determined qualitatively based on the presence of bending angles over 45°. Quantitatively, tortuosity was determined as a distance coefficient and a measure of curvature. In three studies out of six, verification of estimates was carried out by comparing the results with expert opinions (accuracy was 0.92–0.94). The study limitations are as follows: monocentricity, the use of data from one type of equipment.