Patient-specific Channelized Hotelling observer to estimate lesion detectability in CT

Area of Science:

• Medical Imaging
• Radiology
• Image Analysis

Background:

• Task-based image quality assessment is vital for CT protocol and radiation dose optimization.
• A need exists to measure and monitor image quality for individual patient CT exams.
• Current methods may not adequately address patient-specific variations in image quality.

Purpose of the Study:

• To develop and validate a patient-specific method for estimating lesion detectability in CT scans.
• To assess the impact of radiation dose reduction on image quality and lesion detection.
• To provide a tool for monitoring image quality on a per-patient basis.

Main Methods:

• Developed a patient-specific channelized Hotelling observer (CHO)-based method.
• Created an ensemble of background regions from patient images, including uniform and anatomically varying areas.
• Modeled lesion signals incorporating contrast-dependent spatial resolution and estimated index of detectability (d').

Main Results:

• Applied the CHO method to clinical CT images acquired at 100%, 50%, and 25% of standard radiation dose.
• Calculated d' for 5 lesion size/contrast conditions across patient scan ranges.
• Observed average noise levels of 13.2, 17.1, and 21.9, with corresponding average d' values of 3.78, 2.93, and 2.43 at the three dose levels.

Conclusions:

• The developed patient-specific CHO method effectively estimates lesion detectability in CT images.
• The method demonstrates that reducing radiation dose impacts lesion detectability, with lower doses resulting in lower d' values.
• This approach offers a valuable tool for optimizing CT protocols and radiation dose while monitoring patient-specific image quality.