JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Channelized hotelling observer-based low-contrast detectability on the ACR CT accreditation phantom: Part II. Repeatability study

  • 0Department of Radiology, Mayo Clinic, Rochester, Minnesota, USA.
Medical Physics +

|

February 2, 2024

|

February 2, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

Channelized Hotelling observer (CHO) provides repeatable low-contrast detectability measurements for CT quality control. This method is practical for routine use, even with advanced imaging techniques.

Area Of Science

  • Medical Imaging Physics
  • Radiology
  • Image Analysis

Background

  • Routine CT quality control lacks objective, quantitative evaluation for low-contrast detectability correlating with human performance.
  • Channelized Hotelling observer (CHO) is a promising tool but previously considered impractical due to high scan repetition requirements.
  • Previous work optimized CHO for the ACR CT phantom, enabling accurate detectability index (d') measurement with only 1-3 repeat scans.

Purpose Of The Study

  • To validate the repeatability of a CHO-based low-contrast evaluation method.
  • To assess the method's performance across four different CT scanner models.
  • To utilize the American College of Radiology (ACR) CT accreditation phantom for standardization.

Main Methods

  • Repeatability testing on four CT scanners (Siemens Force/Alpha, Canon Prism/Prime SP).
  • Acquisition of 10 phantom repositionings, with 3 scans per repositioning at 24, 12, and 6 mGy.
  • Application of CHO to assess low-contrast object detectability (4-6 mm) and evaluation of repeatability using P<15%.

Main Results

  • CHO measurements demonstrated high repeatability (P<15% of 98.8%-99.9%) at 12 mGy with iterative reconstruction (IR) on all tested scanners.
  • Repeatability remained high across different dose levels and object sizes on individual scanners (P<15% ranging from 91.5% to 99.998%).
  • High repeatability was also observed for deep learning reconstructions and photon-counting-detector (PCD)-CT modes (P<15% > 96.5%).

Conclusions

  • CHO offers highly repeatable measurements for low-contrast detectability on the ACR phantom, exceeding 95% probability within ±15% tolerance.
  • Repeatability is maintained with advanced techniques like deep learning reconstruction and PCD-CT.
  • The study confirms the feasibility of practical CHO implementation for routine CT quality control and performance evaluation.

Keywords: