efg’s R Notes: EBImage: Features
Basic | Haralick | Moment | Shape
library(EBImage)Image Object Features
- computeFeatures
- computeFeatures.basic
- computeFeatures.haralick
- computeFeatures.moment
- computeFeatures.shape
- standardExpandRef
load and segment nucleus
par(mfrow=c(1,2))
y <- readImage(system.file("images", "nuclei.tif", package="EBImage"))[,,1]
x <- thresh(y, 10, 10, 0.05)
x <- opening(x, makeBrush(5, shape='disc'))
x <- bwlabel(x)
display(y, title="Cell nuclei")
display(x, title="Segmented nuclei")
compute shape features
fts <- computeFeatures.shape(x)
str(fts)## num [1:76, 1:6] 75 286 211 321 502 423 379 330 488 204 ...
## - attr(*, "dimnames")=List of 2
## ..$ : chr [1:76] "1" "2" "3" "4" ...
## ..$ : chr [1:6] "s.area" "s.perimeter" "s.radius.mean" "s.radius.sd" ...fts## s.area s.perimeter s.radius.mean s.radius.sd s.radius.min s.radius.max
## 1 75 40 5.384756 2.5473087 1.4630875 9.598991
## 2 286 95 11.914279 6.2510079 1.6594562 22.733608
## 3 211 61 8.463546 2.8625585 3.9930738 12.369067
## 4 321 56 9.781896 1.6472487 6.2744674 12.269780
## 5 502 80 13.182665 4.1268371 6.3981687 19.601315
## 6 423 70 11.193220 2.0760003 5.6671268 14.562252
## 7 379 62 10.640572 1.6783868 7.8950748 13.529297
## 8 330 61 9.992652 1.7385299 6.7247892 13.441793
## 9 488 81 12.307053 3.3772075 6.2079250 18.124605
## 10 204 60 7.874377 2.8414135 2.3777908 12.207398
## 11 44 19 3.332189 0.4755575 2.3139945 4.095806
## 12 311 67 9.494065 3.0864776 1.5623237 13.654335
## 13 44 19 3.305533 0.3703584 2.5004155 3.864749
## 14 420 67 11.123299 1.2934341 8.6474474 13.257580
## 15 389 62 10.703034 1.1018437 8.0865505 12.705296
## 16 621 87 13.992169 2.7306336 8.2129473 18.503919
## 17 361 75 10.111333 3.3552681 2.9636314 14.539479
## 18 357 64 10.244403 1.8122203 7.0146679 13.037899
## 19 435 69 11.305686 1.1548827 8.8329856 13.834683
## 20 129 36 5.980610 0.9313067 4.1970191 7.498405
## 21 298 58 9.361134 2.0215001 4.6039648 12.788327
## 22 193 55 8.004665 2.7831936 3.3462450 13.494008
## 23 518 77 12.850968 1.7065512 9.7696871 16.372055
## 24 375 66 10.605399 1.6752607 6.0521153 13.707692
## 25 296 60 9.568764 2.7346801 4.1162618 13.240479
## 26 293 59 9.326125 2.1519705 4.5920944 12.679326
## 27 319 59 9.592548 1.5283310 5.2950031 11.936404
## 28 298 53 9.330598 0.8379296 7.9462864 11.174941
## 29 41 24 3.425799 1.3888233 1.3076962 5.631169
## 30 420 71 11.239198 2.0212645 7.1682393 15.389424
## 31 426 71 11.499226 3.2205048 4.3185267 16.412005
## 32 527 78 12.593632 2.7746344 7.2046491 17.092787
## 33 320 65 9.968671 3.3952696 3.6402744 14.971436
## 34 394 63 10.875879 1.6449716 7.5135722 13.660037
## 35 207 46 8.142217 2.1090645 4.7037973 11.353780
## 36 225 49 8.116555 1.8312625 4.9458964 11.162650
## 37 397 60 10.974379 1.5693077 7.5465408 13.719218
## 38 136 36 6.157999 0.7461044 4.4466141 7.640909
## 39 348 65 10.110098 2.8345070 3.4814810 13.567975
## 40 281 54 9.075496 1.5849007 6.3138101 11.906011
## 41 376 70 11.071847 2.7499368 6.8124286 16.161967
## 42 75 26 4.486989 0.7050572 2.7846791 5.607740
## 43 429 72 11.359828 2.5598612 5.2518552 15.312442
## 44 436 74 11.583097 3.0002508 4.9133656 16.382442
## 45 111 56 7.362079 3.6954862 1.5642082 13.581117
## 46 464 70 11.963935 2.3913054 6.9746626 15.853442
## 47 485 76 12.036427 2.2339630 7.8179711 15.559382
## 48 315 60 9.620033 1.9466415 5.0364229 12.744890
## 49 320 56 9.732844 1.2712755 7.3639012 11.877886
## 50 368 70 10.636651 2.4402276 5.2173963 14.980695
## 51 390 79 11.033786 3.5889820 1.6093382 15.952305
## 52 390 93 10.405770 4.0866145 1.3026739 16.318674
## 53 319 64 9.871016 2.7185802 5.0637055 14.532846
## 54 339 59 9.941373 0.9332251 8.2734958 11.655066
## 55 536 77 12.687949 2.3561859 8.2420786 16.448169
## 56 342 66 10.049837 2.9300224 2.8931460 14.155221
## 57 507 81 12.527034 3.2253364 6.2175786 17.870011
## 58 248 53 8.552597 1.7548511 4.2934790 11.441401
## 59 345 69 10.565709 3.4667560 3.1461295 15.804499
## 60 351 65 10.169064 2.1134302 5.2726154 13.820292
## 61 421 73 11.224041 3.0790975 3.9368729 16.263845
## 62 358 60 10.230394 0.7600282 8.7451733 11.632342
## 63 364 63 10.597760 2.2577422 7.1604897 15.149611
## 64 207 49 7.727419 1.1778694 5.2080817 10.220110
## 65 380 82 11.863925 5.1505547 2.0893558 19.863548
## 66 582 85 13.019464 2.5345198 7.0551853 16.635515
## 67 309 58 9.500583 1.4418692 6.3927166 11.827988
## 68 408 87 11.064216 4.1956813 0.5892416 16.864643
## 69 273 54 8.889901 1.4330635 5.7806557 11.055804
## 70 315 75 8.611755 3.1603344 0.5259489 12.309750
## 71 352 82 10.428098 4.4000913 2.2639941 17.451083
## 72 400 66 11.139322 2.3482859 7.3767342 15.324505
## 73 262 53 8.669239 0.9545777 6.9192824 10.276785
## 74 345 68 10.415186 3.3805793 3.2596012 15.656468
## 75 321 74 10.503623 3.7010085 4.3515402 17.611904
## 76 16 12 1.833022 0.6018740 0.9718253 2.635231Put heatmap of correlations here.
Features computed by computeFeatures.shape are prefixed with “s.” and all are measured in pixels.
compute features
ft <- computeFeatures(x, y, xname="nucleus")
cat("median features are:\n")## median features are:apply(ft, 2, median)## nucleus.0.m.cx nucleus.0.m.cy
## 2.760080e+02 2.917190e+02
## nucleus.0.m.majoraxis nucleus.0.m.eccentricity
## 2.716022e+01 7.799645e-01
## nucleus.0.m.theta nucleus.0.s.area
## 6.983438e-02 3.450000e+02
## nucleus.0.s.perimeter nucleus.0.s.radius.mean
## 6.450000e+01 1.019973e+01
## nucleus.0.s.radius.sd nucleus.0.s.radius.min
## 2.245853e+00 5.135894e+00
## nucleus.0.s.radius.max nucleus.a.b.mean
## 1.365719e+01 5.537696e-01
## nucleus.a.b.sd nucleus.a.b.mad
## 1.184311e-01 1.162824e-01
## nucleus.a.b.q001 nucleus.a.b.q005
## 3.151373e-01 3.598039e-01
## nucleus.a.b.q05 nucleus.a.b.q095
## 5.647059e-01 7.411765e-01
## nucleus.a.b.q099 nucleus.a.m.cx
## 7.681569e-01 2.759344e+02
## nucleus.a.m.cy nucleus.a.m.majoraxis
## 2.918791e+02 2.563950e+01
## nucleus.a.m.eccentricity nucleus.a.m.theta
## 7.673224e-01 2.233263e-02
## nucleus.a.h.asm.s1 nucleus.a.h.con.s1
## 2.238844e-02 3.645172e+00
## nucleus.a.h.cor.s1 nucleus.a.h.var.s1
## 8.512804e-01 1.240606e+01
## nucleus.a.h.idm.s1 nucleus.a.h.sav.s1
## 5.039755e-01 3.630086e+01
## nucleus.a.h.sva.s1 nucleus.a.h.sen.s1
## 1.269986e+03 1.339942e+00
## nucleus.a.h.ent.s1 nucleus.a.h.dva.s1
## 1.816435e+00 3.645172e+00
## nucleus.a.h.den.s1 nucleus.a.h.f12.s1
## 6.324399e-01 2.786244e-01
## nucleus.a.h.f13.s1 nucleus.a.h.asm.s2
## 6.659335e-01 1.982790e-02
## nucleus.a.h.con.s2 nucleus.a.h.cor.s2
## 9.450356e+00 6.010386e-01
## nucleus.a.h.var.s2 nucleus.a.h.idm.s2
## 1.226194e+01 3.643872e-01
## nucleus.a.h.sav.s2 nucleus.a.h.sva.s2
## 3.686993e+01 1.302210e+03
## nucleus.a.h.sen.s2 nucleus.a.h.ent.s2
## 1.294284e+00 1.858567e+00
## nucleus.a.h.dva.s2 nucleus.a.h.den.s2
## 9.450356e+00 8.014289e-01
## nucleus.a.h.f12.s2 nucleus.a.h.f13.s2
## 1.731739e-01 5.477548e-01
## nucleus.Ba.b.mean nucleus.Ba.b.sd
## 6.112350e-01 2.347311e-01
## nucleus.Ba.b.mad nucleus.Ba.b.q001
## 2.489820e-01 1.576757e-01
## nucleus.Ba.b.q005 nucleus.Ba.b.q05
## 2.603433e-01 6.334505e-01
## nucleus.Ba.b.q095 nucleus.Ba.b.q099
## 1.033540e+00 1.132751e+00
## nucleus.Ba.m.cx nucleus.Ba.m.cy
## 2.759874e+02 2.918643e+02
## nucleus.Ba.m.majoraxis nucleus.Ba.m.eccentricity
## 2.727524e+01 7.903687e-01
## nucleus.Ba.m.theta nucleus.Ba.h.asm.s1
## 8.621972e-03 1.065775e-02
## nucleus.Ba.h.con.s1 nucleus.Ba.h.cor.s1
## 2.245891e+01 7.043244e-01
## nucleus.Ba.h.var.s1 nucleus.Ba.h.idm.s1
## 4.250613e+01 3.103933e-01
## nucleus.Ba.h.sav.s1 nucleus.Ba.h.sva.s1
## 4.025579e+01 1.637727e+03
## nucleus.Ba.h.sen.s1 nucleus.Ba.h.ent.s1
## 1.568549e+00 2.330354e+00
## nucleus.Ba.h.dva.s1 nucleus.Ba.h.den.s1
## 2.245891e+01 9.647075e-01
## nucleus.Ba.h.f12.s1 nucleus.Ba.h.f13.s1
## 1.800243e-01 5.919088e-01
## nucleus.Ba.h.asm.s2 nucleus.Ba.h.con.s2
## 9.993056e-03 5.253647e+01
## nucleus.Ba.h.cor.s2 nucleus.Ba.h.var.s2
## 3.220357e-01 4.363279e+01
## nucleus.Ba.h.idm.s2 nucleus.Ba.h.sav.s2
## 2.085173e-01 4.089912e+01
## nucleus.Ba.h.sva.s2 nucleus.Ba.h.sen.s2
## 1.646988e+03 1.489651e+00
## nucleus.Ba.h.ent.s2 nucleus.Ba.h.dva.s2
## 2.186038e+00 5.253647e+01
## nucleus.Ba.h.den.s2 nucleus.Ba.h.f12.s2
## 1.126337e+00 1.735806e-01
## nucleus.Ba.h.f13.s2
## 5.638763e-01The function computeFeatures.basic computes spatial-independent statistics on pixel intensities that have a “b.” prefix: b.mean, b.sd, b.mad, b.q*. Intensity values range from 0.0 to 1.0.
The function computeFeatures.moment computes features related to object image moments, which can be computed with or without reference intensities, and have a prefix “m.”:
- center of mass: m.cx, m.cy [pixels]
- elliptical fit major axis: m.majoraxis [pixels]
- elliptical eccentricity: m.eccentricity [dimensionless] (0=circle, 1=straight line)
- object angle [radians]: m.theta
The function computeFeatures.haralick computes features that quantify pixel texture. Features are named according to Haralick’s original paper, Textural Features for Image Classification. Haralick texture features have a prefix “h.”.
Texture features in EBImage are calculated according to a Gray-Level Co-occurrence Matrix (GLCM) using the mean intensity of the pixel’s 4 neighboring pixels (0°, 45°, 90°, and 135°) according to Aleksi Räsänen et al: The GLCM features estimate different features of pixel brightness value combinations from the neighboring pixels. Overall, 13 GLCM features given by Haralick et al. (1973) were calculated: angular second moment (asm), contrast (con), correlation (cor), variance (var), inverse difference moment (idm), sum average (sav), sum variance (sva), sum entropy (sen), entropy (ent), difference variance (dva), difference entropy (den), and two measures of correlation (f12 and f13). For the GLCM calculations, the pixel brightness values were quantized to 16 gray levels using equal intervals.
compute feature properties
ftp <- computeFeatures(x, y, properties=TRUE, xname="nucleus")
ftp## name translation.invariant rotation.invariant
## 1 nucleus.0.m.cx FALSE TRUE
## 2 nucleus.0.m.cy FALSE TRUE
## 3 nucleus.0.m.majoraxis TRUE TRUE
## 4 nucleus.0.m.eccentricity TRUE TRUE
## 5 nucleus.0.m.theta TRUE FALSE
## 6 nucleus.0.s.area TRUE TRUE
## 7 nucleus.0.s.perimeter TRUE TRUE
## 8 nucleus.0.s.radius.mean TRUE TRUE
## 9 nucleus.0.s.radius.sd TRUE TRUE
## 10 nucleus.0.s.radius.min TRUE TRUE
## 11 nucleus.0.s.radius.max TRUE TRUE
## 12 nucleus.a.b.mean TRUE TRUE
## 13 nucleus.a.b.sd TRUE TRUE
## 14 nucleus.a.b.mad TRUE TRUE
## 15 nucleus.a.b.q001 TRUE TRUE
## 16 nucleus.a.b.q005 TRUE TRUE
## 17 nucleus.a.b.q05 TRUE TRUE
## 18 nucleus.a.b.q095 TRUE TRUE
## 19 nucleus.a.b.q099 TRUE TRUE
## 20 nucleus.a.m.cx FALSE TRUE
## 21 nucleus.a.m.cy FALSE TRUE
## 22 nucleus.a.m.majoraxis TRUE TRUE
## 23 nucleus.a.m.eccentricity TRUE TRUE
## 24 nucleus.a.m.theta TRUE FALSE
## 25 nucleus.a.h.asm.s1 TRUE TRUE
## 26 nucleus.a.h.con.s1 TRUE TRUE
## 27 nucleus.a.h.cor.s1 TRUE TRUE
## 28 nucleus.a.h.var.s1 TRUE TRUE
## 29 nucleus.a.h.idm.s1 TRUE TRUE
## 30 nucleus.a.h.sav.s1 TRUE TRUE
## 31 nucleus.a.h.sva.s1 TRUE TRUE
## 32 nucleus.a.h.sen.s1 TRUE TRUE
## 33 nucleus.a.h.ent.s1 TRUE TRUE
## 34 nucleus.a.h.dva.s1 TRUE TRUE
## 35 nucleus.a.h.den.s1 TRUE TRUE
## 36 nucleus.a.h.f12.s1 TRUE TRUE
## 37 nucleus.a.h.f13.s1 TRUE TRUE
## 38 nucleus.a.h.asm.s2 TRUE TRUE
## 39 nucleus.a.h.con.s2 TRUE TRUE
## 40 nucleus.a.h.cor.s2 TRUE TRUE
## 41 nucleus.a.h.var.s2 TRUE TRUE
## 42 nucleus.a.h.idm.s2 TRUE TRUE
## 43 nucleus.a.h.sav.s2 TRUE TRUE
## 44 nucleus.a.h.sva.s2 TRUE TRUE
## 45 nucleus.a.h.sen.s2 TRUE TRUE
## 46 nucleus.a.h.ent.s2 TRUE TRUE
## 47 nucleus.a.h.dva.s2 TRUE TRUE
## 48 nucleus.a.h.den.s2 TRUE TRUE
## 49 nucleus.a.h.f12.s2 TRUE TRUE
## 50 nucleus.a.h.f13.s2 TRUE TRUE
## 51 nucleus.Ba.b.mean TRUE TRUE
## 52 nucleus.Ba.b.sd TRUE TRUE
## 53 nucleus.Ba.b.mad TRUE TRUE
## 54 nucleus.Ba.b.q001 TRUE TRUE
## 55 nucleus.Ba.b.q005 TRUE TRUE
## 56 nucleus.Ba.b.q05 TRUE TRUE
## 57 nucleus.Ba.b.q095 TRUE TRUE
## 58 nucleus.Ba.b.q099 TRUE TRUE
## 59 nucleus.Ba.m.cx FALSE TRUE
## 60 nucleus.Ba.m.cy FALSE TRUE
## 61 nucleus.Ba.m.majoraxis TRUE TRUE
## 62 nucleus.Ba.m.eccentricity TRUE TRUE
## 63 nucleus.Ba.m.theta TRUE FALSE
## 64 nucleus.Ba.h.asm.s1 TRUE TRUE
## 65 nucleus.Ba.h.con.s1 TRUE TRUE
## 66 nucleus.Ba.h.cor.s1 TRUE TRUE
## 67 nucleus.Ba.h.var.s1 TRUE TRUE
## 68 nucleus.Ba.h.idm.s1 TRUE TRUE
## 69 nucleus.Ba.h.sav.s1 TRUE TRUE
## 70 nucleus.Ba.h.sva.s1 TRUE TRUE
## 71 nucleus.Ba.h.sen.s1 TRUE TRUE
## 72 nucleus.Ba.h.ent.s1 TRUE TRUE
## 73 nucleus.Ba.h.dva.s1 TRUE TRUE
## 74 nucleus.Ba.h.den.s1 TRUE TRUE
## 75 nucleus.Ba.h.f12.s1 TRUE TRUE
## 76 nucleus.Ba.h.f13.s1 TRUE TRUE
## 77 nucleus.Ba.h.asm.s2 TRUE TRUE
## 78 nucleus.Ba.h.con.s2 TRUE TRUE
## 79 nucleus.Ba.h.cor.s2 TRUE TRUE
## 80 nucleus.Ba.h.var.s2 TRUE TRUE
## 81 nucleus.Ba.h.idm.s2 TRUE TRUE
## 82 nucleus.Ba.h.sav.s2 TRUE TRUE
## 83 nucleus.Ba.h.sva.s2 TRUE TRUE
## 84 nucleus.Ba.h.sen.s2 TRUE TRUE
## 85 nucleus.Ba.h.ent.s2 TRUE TRUE
## 86 nucleus.Ba.h.dva.s2 TRUE TRUE
## 87 nucleus.Ba.h.den.s2 TRUE TRUE
## 88 nucleus.Ba.h.f12.s2 TRUE TRUE
## 89 nucleus.Ba.h.f13.s2 TRUE TRUERelated: Slides 13-22 in Wolfgang Huber’s presentation Machine learning in high-throughput screening and automated phenotyping [~35 MB].
packageVersion("EBImage")## [1] '4.8.2'efg
2015-01-19 1926