Numerical observer for atherosclerotic plaque classification in spectral computed tomography

Area of Science:

• Medical Imaging Physics
• Radiology
• Computational Imaging

Background:

• Spectral computed tomography (SCT) provides enhanced image quality over conventional computed tomography (CT), addressing limitations in atherosclerotic plaque imaging.
• Objective image assessment for discriminating plaques using SCT remains limited in current literature.
• Differentiating vulnerable plaque features is crucial for cardiovascular disease management.

Purpose of the Study:

• To develop and evaluate a numerical-observer method for objective performance assessment in discriminating vulnerable plaque features.
• To compare the plaque discrimination performance of multi-energy CT (MECT), dual-energy CT (DECT), and conventional CT.
• To incorporate spectral information and signal variability into the observer model for realistic clinical task simulation.

Main Methods:

• A two-stage numerical observer was developed, utilizing localized channelized Hotelling observers (CHO) and a final Hotelling observer integrating spectral information.
• The observer incorporated localized prewhitening and matched filtering with Laguerre-Gaussian channel functions for decorrelation.
• Task-based frameworks (SKE/BKE and SKS/BKE) were applied to simulated carotid artery atherosclerosis data, including signal variability, for performance evaluation using SNR and AUC metrics.

Main Results:

• MECT demonstrated superior performance over DECT and conventional CT in all discrimination tasks (calcified vs. fatty-mixed plaque, calcified vs. iodine-mixed blood).
• MECT achieved significant SNR improvements ranging from 46.8% to 67.7% compared to other methods, depending on image reconstruction.
• The proposed numerical observer and signal variability framework effectively assessed material characterization using SCT's energy-dependent attenuation.

Conclusions:

• The developed numerical observer provides a robust method for objective assessment of SCT performance in plaque characterization.
• MECT offers significant advantages over DECT and conventional CT for identifying vulnerable plaque features, improving diagnostic accuracy.
• This approach holds promise for extending to other clinical applications, such as kidney or urinary stone identification.