Pairwise comparison versus Likert scale for biomedical image assessment.

Area of Science:

• Medical imaging
• Radiology
• Image analysis

Background:

• Biomedical imaging research heavily relies on subjective reader analysis.
• Current assessment methods suffer from interreader variability and fixed limits.
• Improved quantitative analysis is crucial for reliable biomedical image interpretation.

Purpose of the Study:

• To compare the performance of pairwise comparison and Likert scale methods for biomedical image analysis.
• To evaluate which method provides more accurate reader assessments of image sharpness.
• To address limitations of current subjective and semi-quantitative image evaluation techniques.

Main Methods:

• A set of 10 digitally blurred chest radiographs was used.
• Readers assessed image sharpness using pairwise comparison and a 10-point Likert scale.
• Lin concordance correlation coefficient (CCC) calculated reader agreement with actual sharpness.

Main Results:

• Pairwise comparison achieved the highest accuracy (CCC, 1.0).
• Ranked Likert scores showed high accuracy (CCC, 0.99), outperforming nonranked Likert scores (CCC, 0.83).
• Accuracy slightly improved with repeated assessments or reference images.

Conclusions:

• Pairwise comparison and ranked Likert scores provide more accurate reader assessments than nonranked Likert scores.
• These findings suggest improved methods for quantitative analysis in biomedical imaging.
• The study highlights the potential to reduce interreader variability in image interpretation.