Image reconstruction techniques using deep learning quality segmentation

Abstract

Translational CT (TCT), in developing nations, a low-end computed tomography (CT) technology are relatively common. The limited-angle TCT scanning mode is often used with large-angle scanning to scan items within a narrow angular range, reduce X-ray radiation, scan long objects, and prevent detector discrepancies.. However, this scanning mode greatly reduces the picture quality and diagnostic accuracy due to the added noise and limited-angle distortions. A U-net convolutional neural network-based approach for limited-angle TCT image reconstruction has been created to reconstruct a high-quality image for the limited-angle TCT scanning mode (CNN). The limited-angle TCT projection data are first examined using the SART method, and the resulting picture is then fed into a trained CNN that can reduce artifacts and maintain structures to provide a better reconstructed image. Simulated studies are used to demonstrate the effectiveness of the algorithm designed for the limitedangle TCT scanning mode. In contrast to certain modern techniques, the developed algorithm considerably lowers noise and limited-angle artifacts while maintaining image structures.

Keywords

• deep learning
• network segmentation
• image reorganisation
• cnn