In medical imaging, constructing 3D representations from 2D images has become crucial for enhanced diagnostic and treatment planning, especially in areas like virtual reality and precision surgery. This research explores methodologies for 2D-to-3D image conversion, emphasizing a specific approach that uses deep learning techniques, particularly the U-Net model, for segmenting critical medical images. The model is initially trained on a large, diverse dataset and then employs transfer learning to accurately segment lung CT scans, compensating for limited data availability. The segmented images undergo depth estimation to capture spatial details, which are then used to construct a 3D point cloud—a precise visual representation of infected areas, valuable for clinical analysis and planning. The field of 2D-to-3D image construction spans both traditional stereoscopic techniques, which rely on binocular vision to estimate depth, and advanced methods using convolutional neural networks (CNNs) for automated depth estimation. While these approaches hold promise, challenges such as managing occlusions and achieving real-time performance remain. Leveraging architectures like deep neural networks and transfer learning techniques has proven essential in overcoming these obstacles. As advancements continue, these methodologies are poised to revolutionize medical imaging, providing clinicians with more accurate, immersive, and actionable 3D representations of critical anatomical structures.