In the present work we carried out a retrospective study on DICOM Metadata stored in two PACS archives of two different health institutions (Institution A and Institution B).

It was requested authorization to the Hospital Board of Directors and to the Ethics Commission, making sure that would be guaranteed the confidentiality of the data collected.

DICOM metadata were indexed and extracted after Dicoogle user validation. Retrieved tag-value data elements were used to objectively assess X-ray exposure.

Due to its generalized availability, the Sensitivity DICOM attribute was chosen as the main exposure related parameter to instance our approach in the mammographic multi-site study spanning 2008-2011 data. Both institutions have Fuji CR systems in which a higher sensitivity value (S Value) is associated with a lower detector exposure to radiation [10-12].

Along with the Sensitivity DICOM attribute, complementary attributes such as Modality, Study Date, Study Description, Sensitivity and Acquisition Device Processing were also collected.

Descriptive statistical tests were used to characterize the average, median and standard deviation of Sensitivity values during the study period.

The Kruskal-Wallis statistical test was used to characterize exposure variation patterns both for intra-site and inter-site assessments. In this test the following hypotheses were considered:

• H0 - There are no differences between the Sensitivity median values over the study period and under the same mammographic projection performed in each health institution;
• H1 - There are differences between the Sensitivity median values over the study period and under the same mammographic projection performed in each health institution.