Hybrid phantom for lung CT: Design and validation

Area of Science:

• Medical Imaging Physics
• Radiological Phantom Development
• Computed Tomography (CT) Protocol Optimization

Background:

• Optimization of CT lung imaging protocols is crucial, particularly with the advent of low-dose CT (LDCT) for lung cancer screening.
• The increasing use of non-linear reconstruction algorithms necessitates phantoms for task-based evaluation, acceptance, and quality control (QC).

Purpose of the Study:

• To present and validate a novel hybrid phantom for lung CT imaging.
• The phantom integrates two distinct setups: one for task-based image quality metrics and an anthropomorphic one for clinical relevance.

Main Methods:

• A hybrid phantom was designed, combining a task-based setup (Mercury) and an anthropomorphic setup (Freddie), mimicking chest structures and lung nodules.
• Validation involved assessing material attenuation properties (Hounsfield Units), a reader study with radiologists and non-radiologists, and task-based metric measurements.
• Images were acquired using clinical thorax protocols with automatic tube current modulation (TCM), two filter combinations, and reconstructed with a deep learning algorithm.

Main Results:

• Nominal and observed Hounsfield Units (HU) showed agreement within 15% for most materials in the task-based setup.
• Reader studies indicated good performance in detecting synthetic nodules and ground-glass opacities (GGO), with scores varying by material and reader group.
• Task-based metrics demonstrated sensitivity to variations in dose, voltage, and filtration, highlighting protocol-dependent detectability index changes.

Conclusions:

• The hybrid phantom uniquely combines task-based (Mercury) and patient-based (Freddie) setups, enhancing its utility for CT protocol optimization.
• This design facilitates the application of the detectability index for optimizing CT protocols in clinical settings, particularly for lung cancer screening.