Simulation of CT image formation and reconstruction using PHITS and radon transform

Purpose: Computed tomography (CT) image formation relies on X-ray attenuation through matter, where projections acquired at different angles represent line integrals used to reconstruct cross-sectional images. Accurate modeling of this process is essential for improving image quality in CT imaging. While Monte Carlo methods are widely used for radiation transport and dosimetry, their application to simulating CT image formation and reconstruction remains limited, particularly using the Particle and Heavy Ion Transport code System (PHITS). This study investigates the capability of PHITS to simulate CT projection data and reconstruct tomographic images using a Radon transform-based approach. Methods: X-ray attenuation through a human phantom in a CT configuration were simulated in PHITS. A stylized phantom representing the thoracic region was defined, and an X-ray source at 120 kV was used to generate projections across multiple angular positions. Projection data were obtained using the [T-Track] tally to represent the detector signal for each angle. The resulting two-dimensional data were then processed in Python, where values were summed along the z-direction to obtain one-dimensional detector profiles. These profiles correspond to projection data consistent with the Radon transform and were combined to form sinograms. Image reconstruction was performed using backprojection, where projection data from all angles were integrated to generate cross-sectional slices. Results: PHITS-generated projections produced coherent sinograms and reconstructible CT images with identifiable anatomical features of the chest phantom. A total of 20 reconstructed slices were generated, demonstrating structural variation along the z-direction of the thoracic region. The reconstructed images showed clear differentiation between major structures such as the lungs, heart, and spine. Blurring and streak artifacts were observed due to limited angular sampling and the absence of filtering. However, the overall anatomical structure was preserved. Conclusion: This demonstrates that PHITS can be used to simulate the acquisition of CT projection data. Cross-sectional images were reconstructed from the simulated projections using Radon transform in Python. The results demonstrate that projection data generated using PHITS can produce reconstructible CT images and provide a basis for further studies on image quality, reconstruction techniques, and imaging artifacts.