Skip to main content

Table 3 Intra-class correlation coefficient, within-reader and between-woman SD of MD measures to assess repeatability of MD readings, by image format

From: Mammographic density assessed on paired raw and processed digital images and on paired screen-film and digital images across three mammography systems

  SFM Raw digital Processed digital
Measure subset N obs N women N repeats ICC Within-reader SD Between-women SD N obs N women N repeats ICC Within-reader SD Between-women SD N obs N women N repeats ICC Within-reader SD Between-women SD
Percent mammographic densitya
 All 6659 6418 241 0.94 0.42 1.61 1243 1098 145 0.91 0.51 1.64 5009 4627 394 0.89 0.49 1.40
 Reader 1 (H1, H2, G2, F1) 1886 1818 68 0.96 0.38 1.90 346 298 48 0.97 0.33 1.87 1539 1413 126 0.91 0.46 1.51
 Reader 2 (H1, H2, G2, F1) 2464 2381 83 0.92 0.46 1.55 356 309 47 0.79 0.63 1.24 1545 1430 119 0.88 0.48 1.32
 Reader 3 (H1, H2, G2, F1) 2309 2217 92 0.89 0.41 1.17 541 489 52 0.87 0.53 1.39 1925 1775 150 0.86 0.52 1.31
 Hologicb (H1, H2)        363 316 47 0.86 0.59 1.48 2742 2517 225 0.87 0.48 1.25
 GEb (G2)        590 536 54 0.92 0.54 1.77 1234 1146 88 0.87 0.59 1.49
 Fujib (F1)        290 244 46 0.94 0.39 1.47 1033 951 82 0.94 0.39 1.52
Dense areaa
 All sets, all readers 6842 6589 253 0.94 0.48 1.82 1244 1099 145 0.88 0.71 1.95 5021 4616 393 0.85 0.67 1.59
 Reader 1 (H1, H2, G2, F1) 1963 1888 75 0.95 0.45 2.06 346 298 48 0.97 0.40 2.12 1543 1417 126 0.89 0.59 1.66
 Reader 2 (H1, H2, G2, F1) 2568 2482 86 0.93 0.49 1.81 357 310 47 0.71 0.89 1.39 1549 1426 119 0.85 0.64 1.50
 Reader 3 (H1, H2, G2, F1) 2311 2217 94 0.89 0.48 1.41 541 489 52 0.84 0.75 1.72 1929 1778 151 0.80 0.75 1.47
 Hologicb (H1, H2)        363 316 47 0.84 0.86 1.94 2745 2520 225 0.83 0.64 1.43
 GEb (G2)        591 537 54 0.86 0.75 1.87 1243 1154 89 0.77 0.85 1.53
 Fujib (F1)        290 244 46 0.94 0.44 1.82 1033 951 82 0.94 0.50 1.95
Breast areaa
 All 6597 6357 240 1.00 0.14 2.46 1243 1098 145 1.00 0.15 2.53 5009 4616 393 1.00 0.11 2.76
 Reader 1 (H1, H2, G2, F1) 1873 1805 68 1.00 0.10 2.46 346 298 48 1.00 0.06 2.49 1539 1413 126 1.00 0.07 2.74
 Reader 2 (H1, H2, G2, F1) 2442 2359 83 0.99 0.18 2.47 356 309 47 1.00 0.17 2.48 1545 1426 119 1.00 0.10 2.77
 Reader 3 (H1, H2, G2, F1) 2282 2191 91 1.00 0.12 2.46 541 489 52 1.00 0.18 2.59 1925 1775 150 1.00 0.13 2.76
 Hologicb (H1, H2)        363 316 47 1.00 0.08 2.13 2742 2517 225 1.00 0.09 2.54
 GEb (G2)        590 536 54 0.99 0.19 2.46 1234 1146 88 1.00 0.09 2.53
 Fujib (F1)        290 244 46 1.00 0.15 2.73 1033 951 82 1.00 0.15 3.01
  1. Analysis: ICCs, within-reader SD and between-women SD were estimated from a one-way ANOVA using all ICMD measurements and sets H1, H2, G2, F1 and F2. Number of repeats is the number of images read at least twice, by the same or different readers. Reader 4 does not appear here because no repeated readings were available for this reader
  2. Numbers of observations vary by MD measure because only dense area was measured if the breast edge was not visible, and only percent mammographic density if the pixel size was unknown
  3. aAnalysed on a square-root scale
  4. bWithin reader, within image type
  5. Obs observations, SFM Screen-film mammography, ICC Intra-class correlation coefficient, MD Mammographic density, SD Standard deviation, N number of, GE General Electric