Development of a separability index for task specific characterization of spectral computed tomography

Area of Science:

• Medical Imaging
• Signal Processing
• Metrology

Background:

• Spectral CT enables material differentiation based on X-ray attenuation.
• Characterizing signal overlap between materials is essential for accurate spectral CT analysis.

Purpose of the Study:

• To develop and evaluate a signal detection-based metrology for quantifying material signal separability in spectral CT.
• To assess the impact of imaging parameters and material concentration on signal separability.

Main Methods:

• Modeled signal response as a random process with deterministic signal and noise.
• Employed a linear Hotelling observer to estimate test statistic distributions for material identification.
• Derived separability index (s') and area under the curve metrics using experimental and simulated photon-counting CT data.

Main Results:

• Optimal conditions for low noise and k-edge signal expression yielded highest separability.
• Material concentration demonstrated the most significant influence on signal separability.
• This impact persisted even under challenging separation conditions in simulated data.

Conclusions:

• The developed metrology effectively measures signal overlap between materials in spectral CT.
• The separability index quantitatively relates image formation factors to material signal responses.
• Task-specific metrology is valuable for optimizing spectral CT acquisition protocols.