Trend of contrast detection threshold with and without localization

Area of Science:

• Medical Imaging
• Radiology
• Human Factors in Imaging

Background:

• Published contrast detection threshold data often relies on location-known methods.
• Previous research on Digital Imaging and Communications in Medicine (DICOM) Grayscale Standard Display Function (GSDF) used a location-unknown approach to mimic clinical practice.
• High variance in prior studies necessitated exploring the impact of location-known versus location-unknown methodologies.

Purpose of the Study:

• To investigate the influence of location-known versus location-unknown methodologies on reader performance in simulated mammographic background analysis.
• To determine the effect of these methodologies on contrast detection thresholds and Reader Operating Characteristic (ROC) curve analysis.

Main Methods:

• Fourteen readers evaluated two sets of simulated mammographic images: one using location-unknown and one using location-known methods.
• Reader performance was analyzed using ROC methodology and a paired t test.
• Contrast detection thresholds were assessed using contingency tables.

Main Results:

• No statistically significant difference was observed in GSDF testing between the two methodologies.
• A highly significant difference (p < 0.0001) was found in the ROC (Area Under the Curve) analysis favoring the location-known methodology.
• The location-known method improved test power and reduced the contrast detection threshold from +3 to +2 gray levels.

Conclusions:

• The choice between location-known and location-unknown experimental designs significantly impacts reader study outcomes, particularly ROC analysis.
• The location-known methodology enhances the sensitivity and accuracy of detecting subtle differences in medical images.
• Careful consideration of methodology is crucial for ensuring experimental conclusions align with research objectives in medical imaging.