Інтелектуальні алгоритми формування та аналізу медичних зображень

• Date: 2026
• Main Authors: Чехместрук, Р.Ю., Сікільчук, К.В., Мартинюк, Т.Б., Майданюк, В.П.
• Format: Article
• Language: Ukrainian
• Published: Vinnytsia National Technical University 2026
• Subjects: медичні зображення; сегментація; реконструкція; глибоке навчання; U-Net; 3D-модель; штучний інтелект; формування зображень
• Online Access: https://oeipt.vntu.edu.ua/index.php/oeipt/article/view/873
• Journal Title: Optoelectronic Information-Power Technologies

The paper considers the application of intelligent algorithms for medical image analysis and formation in tasks of personalized 3D reconstruction of human anatomical structures. Modern medical image segmentation methods are analyzed and their main limitations are identified, including insufficient accuracy in the presence of noise, artifacts, and damaged image regions. A reconstruction method combining deep learning algorithms and anatomical symmetry principles is proposed. An experimental study and comparison with baseline approaches were conducted. The obtained results demonstrate improved reconstruction accuracy and better restoration quality of anatomical structures. The study developed an intelligent method for personalized 3D reconstruction of anatomical structures, which combines deep learning algorithms, spatial analysis and symmetrical restoration of damaged areas. The proposed method is based on the use of the U-Net neural network for segmentation of medical images and the subsequent application of geometric principles of symmetry for the reconstruction of missing parts of the structure. The experimental study confirmed the effectiveness of the proposed method. The results obtained showed an improvement in the Dice and IoU indicators compared to classical algorithms and standard U-Net segmentation. In addition, the use of symmetrical restoration allowed to reduce the reconstruction error and improve the geometric integrity of three-dimensional models.