Mean glandular dose in digital mammography: a dose calculation method comparison

Area of Science:

• Medical Physics
• Radiology
• Radiation Dosimetry

Background:

• Accurate estimation of mean glandular dose (MGD) is crucial for mammography quality assurance and patient safety.
• Digital Imaging and Communication in Medicine (DICOM) headers contain valuable data for dose assessment.
• Existing methods for calculating MGD include those by Dance et al., Wu et al., and Boone et al.

Purpose of the Study:

• To analyze the agreement between organ dose estimates and calculated doses from clinical digital mammography data.
• To investigate potential biases in organ dose estimation across different digital mammography unit manufacturers.

Main Methods:

• Extracted DICOM header information from 52,405 anonymized mammograms.
• Filtered data for specific breast thicknesses (20-110 mm) and excluded cases with breast implants.
• Calculated MGD using Dance, Wu, and Boone methods; employed Bland-Altman analysis and regression to assess agreement.

Main Results:

• Statistically significant bias was observed between organ dose and calculated doses.
• Bias varied by manufacturer: Philips showed the lowest bias (overestimating Dance by 0.03 mGy), while GE exhibited the highest (overestimating Wu by 0.20 mGy).
• Hologic and Fujifilm organ doses underestimated calculated doses by 0.07 mGy and 0.09 mGy, respectively.

Conclusions:

• Organ dose estimates derived from mammography units disagree with calculated doses.
• Manufacturer-specific biases necessitate caution when using organ dose for rapid dose audits.
• Potential over- or underestimation errors must be considered in reported results based on organ dose alone.