A generic framework to simulate realistic lung, liver and renal pathologies in CT imaging

Area of Science:

• Medical Imaging
• Computational Modeling
• Radiology

Background:

• Realistic three-dimensional (3D) mathematical models of subtle lesions are crucial for computed tomography (CT) performance evaluation and optimization.
• Existing methods may lack the fidelity required for advanced CT system development.

Purpose of the Study:

• To develop a generic mathematical framework for creating realistic 3D lesion models.
• To implement a method for generating hybrid CT datasets using these models.
• To validate the fidelity and realism of the generated models and images.

Main Methods:

• Developed a mathematical framework defining lesion characteristics (size, shape, contrast, contrast-profile).
• Created lesion models from real CT data and inserted them into CT images to form hybrid datasets.
• Assessed model goodness-of-fit using coefficient of determination (R(2)) and image realism via observer studies with Receiver Operator Characteristic (ROC) analysis.

Main Results:

• The lesion models demonstrated a good fit to real data, with an average R(2) of 0.80.
• Observer studies showed no significant difference between real and simulated lesions (Area Under ROC Curve = 0.55).
• A library of realistic lesion models and hybrid CT images was successfully created.

Conclusions:

• The developed framework effectively generates realistic 3D lesion models and hybrid CT images.
• These models are valuable tools for the performance evaluation and optimization of novel CT systems.
• The approach enhances the development cycle for advanced medical imaging technologies.