Advancing Medical Imaging with Synthetic Data

Medical imaging is crucial in clinical diagnosis and treatment monitoring as it provides specific information about the human body. Imaging modalities, including Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Positron Emission Tomography (PET), are commonly used in clinical workflow, each providing unique structural, functional, and metabolic information that supports comprehensive clinical decisions. However, specific imaging modalities, such as PET and CT, pose a risk of radiation exposure, particularly for paediatric patients. Furthermore, the acquisition process of comprehensive multi-modal images is costly, and longer scanner time also introduces artefacts. Consequently, acquiring precise imaging in a safer manner is challenging in practical applications. To address these challenges with medical imaging, we introduced deep learning based medical image synthesis methods by enabling the translation of images from one modality to another. This helps in maximizing the utility of acquired images and reduces scanner time and operation costs to make healthcare accessible for everyone.