Comparing the use and interpretation of PGMI scoring to assess the technical quality of screening mammograms in the UK and Norway

• No previous published work comparing PGMI use between different countries.
• Variation in number of assessment criteria used and their interpretation.
• Best agreement was Norway scoring MLO views from both centres-moderate.
• Least agreement was UK raters scoring CC views from both centres-poor.
• Most frequent faults identical in both test sets and were on MLO views.

ObjectivesTo compare PGMI systems used in the UK and Norway, determine levels of agreement in its interpretation for radiographers within and between centres, informing further research towards developing a more quantitative, uniform system.MethodsMammograms from 112 women consecutively screened in the UK and Norway were anonymised, numbered and enriched to include all four PGMI categories. Cases were scored by five mammographers from each centre using local PGMI. Sets were exchanged and the process repeated. Distribution of categories was recorded and faults documented for images scored less than perfect. These were compared within and between centres and agreement analysed using non-weighted kappa statistic.ResultsNorway uses 38 assessment criteria, the UK uses 15. Best agreement was between Norway raters scoring MLO views from both UK(RMLO k = 0.57, LMLO k = 0.490) and Norway (RMLO k = 0.48, LMLO k = 0.470). Least agreement was between UK raters scoring CC views from both UK(RCC k = 0.007, LCC k = 0.01) and Norway(RCC k = −0.04, LCC k = −0.003). There were no other apparent trends in inter-rater assessment. Most frequent faults in both test sets were on MLO views. Two out of three most common faults were the same for UK and Norway raters.ConclusionsUse of PGMI varied between centres in both number and interpretation of criteria employed. We identified the most common mammographic faults highlighting possible training needs. We suggest further work to provide a consensus list of visual criteria with accurate descriptors for each classification category. A validated way of applying them could help to standardise the process.