This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Breast, Mammography, Screening, Quality assurance
Authors:
R. Gullien, J.-G. Andersen, A. E. Haakull, L. Sandvik, P. Skaane; Oslo/NO
DOI:
10.1594/ecr2016/C-0673
Methods and materials
Patient Selection
All digital mammograms (DM) in this study were performed in the same county.
The numbers of screened women in 2007 were 14.329 and 15.612 in 2009.
The number of women with IC in 2007 was 29 (IC rate 20.4) and in 2009 the number of IC was 39 (IC rate 25.1).
Two women had bilateral cancers,
one in 2007 and one in 2009.
Two women (one in 2007 and one in2009) had only one breast,
due to previous mastectomy.
Exclusion criteria
Two women with implants were excluded,
one in each year,
due to unclear PGMI criteria for images with implants.
Women opted out of the NBCSP are excluded.
This gives a total of 66 women,
and 68 MLO images with IC (n = 24 R-MLO; n = 44 L-MLO).
Radiographers score
Two experienced PGMI radiographer retrospective and independently PGMI classified baseline mammograms prior to the cancer diagnosis.
Each image was classified in one of the four categories: P,
G,
M or I.
The classification categories P+G were pooled.
Images with divergent results were discussed in a consensus meeting.
The PGMI radiographers used the images criteria from the quality assurance manual (QAM) for the NBCSP 2003 edition (1).
The criteria in this edition were valid for exams performed in 2007 and 2009.
Radiological score
Three experienced radiologist's retrospective and independently classified the baseline screening mammograms prior to the cancer diagnose.
Each exam was classified in one of the four categories.
The classification categories were: 1 = negative/normal: no suspicious abnormalities; 2 = nonspecific minimal sign: subtle findings,
not requiring workup; 3 = significant minimal sign: minor findings that in which recall might be considered and 4 = overlooked/missed cancer: findings should have been recalled for workup.
Exams with divergent results were discussed in a consensus meeting.
The radiological assessment categories 1+2 and 3+4 were pooled.
Statistical analyses
For analysis,
we selected only those 68 MLO images of women that later presented with interval cancers.
We used Chi-Square statistic to compare differences between the radiographers PGMI score and the radiologist’s classification.