Automatic quantitative low contrast analysis of digital chest phantom radiographs

Area of Science:

• Medical Imaging
• Radiography
• Image Analysis

Background:

• Low contrast object evaluation in phantom images has traditionally been subjective and labor-intensive.
• Manual densitometry measurements are time-consuming and limit quantitative analysis.
• Advancements in digital radiography enable automated detection and computation processes.

Purpose of the Study:

• To develop and examine an automated method for detecting and computing the subject-to-noise ratio (SNR) of low contrast objects.
• To assess the accuracy and consistency of the automated algorithm in evaluating contrast detail phantoms.

Main Methods:

• An algorithm was developed to automatically detect low contrast disks within a geometric chest phantom.
• The algorithm computes the subject-to-noise ratio (SNR) for these detected objects.
• The accuracy of object localization was evaluated to be less than one pixel.

Main Results:

• The automated algorithm successfully detected and located low contrast objects with sub-pixel accuracy.
• The computed subject-to-noise ratio (SNR) results align with established understanding.
• The method demonstrates consistency and reliability for quantitative phantom evaluations.

Conclusions:

• The developed algorithm automates the detection and SNR computation for low contrast objects in phantom images.
• This automated approach significantly simplifies and enhances the quantitative evaluation of contrast detail phantoms.
• The findings support the adoption of automated methods for more efficient and objective radiographic assessments.