This repository contains the source code and documentation for my Bachelor Thesis project, completed over an 11-week period. The project investigates how machine learning can accelerate dual-energy CT (DECT) scan reconstruction.
Project website: https://luukfroling.github.io/BEP/
Dual-energy CT (DECT) imaging acquires scans using two distinct X-ray energy levels, providing material-specific diagnostic information. However, reconstructing high-quality DECT images is computationally expensive.
This project explores how deep learning, specifically an attention-based U-Net, can learn to approximate results from an existing iterative reconstruction algorithm. The goal is to reduce computation time while maintaining comparable image quality.