0 1992 Wiley-Liss, Inc.
Cytometry 13:485-489 (1992)
Influence of Different Cell Extraction Methods on Cytometric Features F.B.J.M. Thunnissen, H. Perdaen, and J. Forrest Department of Pathology, University of Limburg, Maastricht, (F.B.J.M.T.)and Becton Dickinson Image Cytometry Systems, Leiden (J.F.),The Netherlands, and Becton Dickinson, Imagecytometry Systems, Erembodegem, Belgium (H.P.) Received August 7,1991;accepted January 11,1992
The purpose of this study was to investigate the influence of different cell extraction procedures on relative nuclear DNA content (IOD), nuclear area, and texture features of Feulgen-stained nuclei. In imprints and smears of fineneedle aspirates and suspensions from one human liver specimen, 50 diploid, 50 tetraploid, and 25 octaploid nuclei were measured from each slide. In addition, for DNA measurements, the progressive mean of IOD and tetraploidldiploid and octaploidldiploid ratios was calculated. The results show that the progressive mean of the IOD is constant after measuring 25-30 nuclei. For the three types of specimens, the IOD of diploid nuclei varied slightly. The average coefficient of variation was about 5% for the fineneedle aspirates, imprints, and suspen-
In general, quantification of DNA can be done by flow and image cytometry. In studies comparing both techniques, concordant results in 85-90% of cases were obtained (4,5,9,10,22).The precision of flow cytometry is slightly better based on our experience that for a diploid stemline coefficient of variations of 2-3% and 3 4 % are achieved under optimal conditions for flow and image cytometry systems, respectively (3,11,20). The advantages of DNA image cytometry are that the measured cells can be visualised at the moment of measurement and that, with a limited number of cells, trustworthy data can be obtained (20). When flow cytometry is used on paraffin-embedded archive material, reliable internal controls for ploidy calibrations are difficult to obtain, and diploid peaks may show a large coefficient of variation (CV). Occasionally, in some of these broad diploid peaks, aneuploidy may be present when the same population is measured with image cytometry (13). Image cytometric studies in routine pathology allow
sions. For all tissue sampling methods, the 9Wo confidence limits of the tetraploidldiploid ratio and octaploidldiploid ratio were within the range of 1.9-2.1 and 3.743, respectively. The nuclear area and most of the texture features showed a significant difference (p < 0.01) between the three sampling methods in all nuclear populations. In conclusion, different tissue sampling methods have little or no effect on DNA-related IOD measurements, whereas the outcome of nuclear area and texture features is very dependent of the cell extraction procedure. o 1992 Wiley-Liss, Inc. Key terms: Image analysis, cytometry, sampling methods, accuracy, reproducibility, DNA quantitation
use of quantitative features such as area and texture features in addition to amount of DNA. Possible sources of error in absorbance measurements of nuclei are sample handling before preparation, variations in pressure during specimen preparation, fixation, staining, microscope setting such as glare, shading, diffraction and focusing, digitizer, and computation including segmentation (12,16,18,21).For comparison of nuclear populations within the same specimen, the influence of all these possible sources of error will be limited if the populations are measured simultaneously. However, for comparison of nuclear populations between different specimens, variation in specimen preparation, fixation, and staining may have dramatic influences on some features. The aim of this study was to investigate the influence of different tissue sampling procedures Address reprint requests to Dr. Frederic B.J.M.Thunnissen, Department of Pathology, University of Limburg, P.O. Box 1918,6201 BX Maastricht, The Netherlands.
486
THUNNISSEN ET AL.
Mean IOD
l o o \I
8ol 60
inson, Erembodegem, Belgium) in batches of 14 or 15 specimens with 1 or 2 CAS calibration slides. In brief, the specimen was hydrolysed in 5 N HCL for 60 min, placed in thionin containing staining solution for 60 min, and then rinsed using the CAS reagents. After washing in deionized water for 5 min, dehydration in absolute ethanol and xylene, the slides were mounted with entellan. All handling was done at room temperature.
DNA Cytophotometry Determination of the quantitative features was done using the CAS model 200 image analysis system equipped with the quantitative DNA plus software module as described previously for the CAS 100 system (3,6,20). Using a 40x objective pixel size was 0.5 x 0.4 km2 (3). The CAS system is calibrated for mass measurements in picograms using the predeposited control cells staining a known amount of DNA on each slide. 2o From this calibration slide 50 tetraploid rat nuclei X Mean Mean + l x SD were measured as control cells. A total of 27 FNA, 27 imprints, and 14 suspensions specimens were measured. Based on the differences in size and intensity, 0 0 10 20 30 40 60 nuclei were a priori assigned to either diploid, tetraploid, or octaploid category. Nuclei were selected a t Number of cells random, starting in one corner of the slide and scanFIG.1. The progressive mean with standard deviation of the IOD is ning along the abscissa until the end and backward at shown for a number of diploid nuclei of an imprint. The coefficient of a distance of 2 fields of view. In a slide, 51 (range 48variation remained constant (3.9%)after 25 nuclei. 53) diploid, 50 (range 48-52) tetraploid, and 25 (range 23-28) octaploid nuclei were measured. The automatic segmentation option was used. From on cytometric features such as relative nuclear DNA the digitized nuclear image, the following cytometric content (IOD) and nuclear area and texture features. For this purpose, fine-needle aspirates (FNA), im- features were calculated: IOD (amount of DNA), ploidy prints, and smears of suspensions were made from one value, area, and a number of Markov chain grain dependent texture features. These were contrast, inverse normal human liver specimen. difference moment, difference variance, difference entropy, coefficient of variation (3) and blobness (141, and MATERIALS AND METHODS a statistic feature standard deviation (independent of Specimen Preparation Small pieces of one normal human liver obtained a t grain) (3). The DNA 4ct2c ratio and 8c/2c ratio were autopsy within 8 h after death were snap-frozen. After calculated from the tetraploid and octaploid peaks in defrosting, three different types of cytological samples the DNA histogram by dividing the mean IOU value of were prepared on separate albumin-coated slides (19). the tetraploid and octaploid peak by the mean IOD First, FNA were taken from the blocks. Subsequently, value of the diploid peak. The deviation from linearity imprints were prepared with a maximum of 6 imprints was determined by dividing the difference between the of each side. From the same blocks a suspension was measured and hypothetical value by the hypothetical made by gentle mechanical dissociation of the liver in value. The deviation of linearity was expressed in perPBS buffer at 4 "C. After filtering through a 50-km centages. IOD in arbitrary units (AU) for a nuclear population gauze and centrifugation a t 400 g for 5 min, the pellet was resuspended in a similar volume of PBS and in a specimen was obtained by multiplying the IOD smears were spread out. All specimens were air-dried value by 1.87 and dividing by the mean rat tetraploid for 30 min and then fixed in 4% buffered formaldehyde IOD value. The ratio rat tetraploid DNAhuman dipfor 30 min followed by a short rinse in tap water and loid DNA leads to 1.87. then air drying. Specimens were stored a t room temStatistics perature in a closed box until staining. Mean and SD of the cytometric features were deterStaining mined for the diploid, tetraploid, and octaploid nuclei of All specimens were Feulgen stained according to the each specimen, as well as for the type of tissue samCAS quantitative DNA staining kit (Becton and Dick- pling used. In order to establish the minimum number
487
CELL EXTRACTION METHODS AND CYTOMETRIC FEATURES
of nuclei to be measured in a specimen, the progressive mean was calculated (7,171. In each nuclear population, the Kruskal-Wallis test was used t o compare the difference between the 3 sampling methods for diploid, tetraploid, and octaploid nuclei separately. Each of these comparisons was undertaken a t the 0.01 significance level. If an overall difference was found, the Mann-Whitney U test was used for pairwise comparison of the different sampling methods at the 0.003 level. The null hypothesis was that no differences were present. The significant levels were chosen so that the probability of rejecting one of the null hypothesis for any of the features was a t most 0.09.
RESULTS All specimens showed strongly stained round nuclei with sharp demarcation and a few nuclei with inhomogeneous staining, and irregular edges were present. For measurements, only strong stained round nuclei with sharp demarcation were used. An example of the progressive mean and standard deviation of the progressive mean is shown for the IOD of diploid nuclei in an imprint (see Fig. 1). This is typical of all the slides. The progressive mean and standard deviation of the progressive mean were more or less stable for all specimens after measuring 25-30 nuclei. Therefore, a reliable statistic of DNA ploidy and CV for a single stemline with no or little proliferation is obtained after 30 nuclei are measured. Mean and standard deviation of the cytometric features are shown for diploid, octaploid, and tetraploid nuclei measured in FNA, imprints, and suspensions in Table 1. The results of the Kruskal-Wallis and MannWhitney U tests are shown in Table 2. For the 3 types of specimens, the IOD of diploid nuclei varied slightly. These differences were not significant ( p < 0.01). However, a comparison of imprints and suspensions reveals a significant difference for tetraploid and octaploid nuclei. The average coefficient of variation for IOD from each nuclear population remains below 6% in all specimen types. Nuclear area was lowest for imprints and highest in suspensions for the three nuclear types. For all nuclear populations, a significant difference ( p < 0.003) was present between nuclear areas in the 3 tissue sampling methods. These data prove that the feature nuclear area is very sensitive to specimen preparation. Comparison of the tissue sampling methods reveals significant differences within all nuclear populations for the majority of the texture features. For the 3 types of specimens, the ratios between 4cI2c and 8cI2c peaks and deviation from linearity are shown in Table 3. Note that the 4c/2c ratios are close to the theoretical value of 2 and that the deviation from linearity is approximately 3%. The 99%confidence limits were well within the range of 1.9-2.1, irrespective of the type of specimen measured. The 8cI2c ratios were close to the expected value of 4. The deviations of lin-
Table 1 Mean and Standard Deviation (sd) of Integrated Optical Density (IOD), Nuclear Area, and Texture Features (T) for Diploid, Octaploid, and Tetraploid Nuclei in FNA, Imprints, and Suspensions"
Diploid IOD [AU]
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev Tetraploid IOD rAUl
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev Octaploid IOD [AU]
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev
Imprint mean sd
FNA mean sd
Suspension mean sd
0.95 f 0.06 4.7 2 1.4 42.4 f 5.3 48 t 10 0.34 f 0.02 23.3 t 4.8 0.60 t 0.02 0.31 f 0.02 0.25 t 0.02 0.10 f 0.01
1.0 t 0.06 4.8 f 1.8 56.6 t 7.0 25.8 f 7.3 0.40 f 0.03 13.6 2 3.5 0.60 f 0.02 0.33 2 0.01 0.25 0.01 0.07 2 0.01
0.98 f 0.05 5.3 f 1.7 67.1 t 1.9 17 f 1.7 0.42 f 0.01 8.7 2 0.9 0.57 f 0.01 0.35 0.01 0.26 f 0.01 0.06 0.02
1.88 2 0.08 4.6 f 1.2 70.3 t 9.7 44 f 13 0.38 f 0.02 22 f 6 0.55 5 0.02 0.35 2 0.01 0.22 t 0.01 0.09 f 0.01
2.03 0.12 4.7 t 0.8 96 f 17 26 t 6 0.41 f 0.03 13 f 3 0.56 t 0.04 0.36 f 0.03 0.25 0.03 0.07 2 0.01
1.96 t 0.12 5.8 t 1.7 116 f 5.6 18.6 2 2.1 0.45 2 0.01 10 1 0.52 ? 0.02 0.4 f 0.01 0.28 t 0.01 0.07 t 0.001
3.80 2 0.14 4.4 f 1.5 119 11 55 f 18 0.42 2 0.01 29 f 9 0.51 f 0.01 0.40 f 0.01 0.26 0.01 0.13 t 0.02
4.06 f 0.21 4.3 f 1.1 155 t 25 30 f 7 0.43 t 0.03 16 t 3 0.52 f 0.04 0.39 f 0.03 0.26 0.03 0.15 f 0.21
3.92 k 0.08 5.4 t 1.2 207 t 22 18.3 f 1.8 0.49 f 0.02 11.2 f 1.0 0.47 k 0.02 0.45 f 0.02 0.31 +- 0.02 0.08 +- 0.01
* *
* *
*
*
* *
*
9nv = inverse; d = difference; coef var = coefficient of variation; st dev = standard deviation texture.
earity were slightly higher at this end of the density scale (74.9%). For the 3 types of specimens, the 99% confidence limits were within the 3.7-4.3 region.
DISCUSSION This study deals with the variation in cytometric features and accuracy and linearity of DNA image analysis when measured in imprints, FNA and smears of suspensions. The results of the present study on nuclei of one normal human liver specimen show that differences in tissue sampling methods have little or no influence on DNA measurements (see Table 1). For measurements of normal (diploid) cells from several human organs, Taylor et al. (20) showed a high degree of accuracy and reproducibility of DNA content measurements. Peaks in the DNA histogram with a DNA index between 0.9 and 1.1 were regarded as diploid (20). The results of the Taylor study (20) are in complete concordance with ours.
488
THUNNISSEN E T AL.
Table 2 Summary of the Kruskal-Wallis (K-W) and Mann-Whitney U Test"
K-W FNA-IMP FNA-SUS IMP-SUS Dipl 0.03 * ns Tetr
Cytometry 13:485-489 (1992)
Influence of Different Cell Extraction Methods on Cytometric Features F.B.J.M. Thunnissen, H. Perdaen, and J. Forrest Department of Pathology, University of Limburg, Maastricht, (F.B.J.M.T.)and Becton Dickinson Image Cytometry Systems, Leiden (J.F.),The Netherlands, and Becton Dickinson, Imagecytometry Systems, Erembodegem, Belgium (H.P.) Received August 7,1991;accepted January 11,1992
The purpose of this study was to investigate the influence of different cell extraction procedures on relative nuclear DNA content (IOD), nuclear area, and texture features of Feulgen-stained nuclei. In imprints and smears of fineneedle aspirates and suspensions from one human liver specimen, 50 diploid, 50 tetraploid, and 25 octaploid nuclei were measured from each slide. In addition, for DNA measurements, the progressive mean of IOD and tetraploidldiploid and octaploidldiploid ratios was calculated. The results show that the progressive mean of the IOD is constant after measuring 25-30 nuclei. For the three types of specimens, the IOD of diploid nuclei varied slightly. The average coefficient of variation was about 5% for the fineneedle aspirates, imprints, and suspen-
In general, quantification of DNA can be done by flow and image cytometry. In studies comparing both techniques, concordant results in 85-90% of cases were obtained (4,5,9,10,22).The precision of flow cytometry is slightly better based on our experience that for a diploid stemline coefficient of variations of 2-3% and 3 4 % are achieved under optimal conditions for flow and image cytometry systems, respectively (3,11,20). The advantages of DNA image cytometry are that the measured cells can be visualised at the moment of measurement and that, with a limited number of cells, trustworthy data can be obtained (20). When flow cytometry is used on paraffin-embedded archive material, reliable internal controls for ploidy calibrations are difficult to obtain, and diploid peaks may show a large coefficient of variation (CV). Occasionally, in some of these broad diploid peaks, aneuploidy may be present when the same population is measured with image cytometry (13). Image cytometric studies in routine pathology allow
sions. For all tissue sampling methods, the 9Wo confidence limits of the tetraploidldiploid ratio and octaploidldiploid ratio were within the range of 1.9-2.1 and 3.743, respectively. The nuclear area and most of the texture features showed a significant difference (p < 0.01) between the three sampling methods in all nuclear populations. In conclusion, different tissue sampling methods have little or no effect on DNA-related IOD measurements, whereas the outcome of nuclear area and texture features is very dependent of the cell extraction procedure. o 1992 Wiley-Liss, Inc. Key terms: Image analysis, cytometry, sampling methods, accuracy, reproducibility, DNA quantitation
use of quantitative features such as area and texture features in addition to amount of DNA. Possible sources of error in absorbance measurements of nuclei are sample handling before preparation, variations in pressure during specimen preparation, fixation, staining, microscope setting such as glare, shading, diffraction and focusing, digitizer, and computation including segmentation (12,16,18,21).For comparison of nuclear populations within the same specimen, the influence of all these possible sources of error will be limited if the populations are measured simultaneously. However, for comparison of nuclear populations between different specimens, variation in specimen preparation, fixation, and staining may have dramatic influences on some features. The aim of this study was to investigate the influence of different tissue sampling procedures Address reprint requests to Dr. Frederic B.J.M.Thunnissen, Department of Pathology, University of Limburg, P.O. Box 1918,6201 BX Maastricht, The Netherlands.
486
THUNNISSEN ET AL.
Mean IOD
l o o \I
8ol 60
inson, Erembodegem, Belgium) in batches of 14 or 15 specimens with 1 or 2 CAS calibration slides. In brief, the specimen was hydrolysed in 5 N HCL for 60 min, placed in thionin containing staining solution for 60 min, and then rinsed using the CAS reagents. After washing in deionized water for 5 min, dehydration in absolute ethanol and xylene, the slides were mounted with entellan. All handling was done at room temperature.
DNA Cytophotometry Determination of the quantitative features was done using the CAS model 200 image analysis system equipped with the quantitative DNA plus software module as described previously for the CAS 100 system (3,6,20). Using a 40x objective pixel size was 0.5 x 0.4 km2 (3). The CAS system is calibrated for mass measurements in picograms using the predeposited control cells staining a known amount of DNA on each slide. 2o From this calibration slide 50 tetraploid rat nuclei X Mean Mean + l x SD were measured as control cells. A total of 27 FNA, 27 imprints, and 14 suspensions specimens were measured. Based on the differences in size and intensity, 0 0 10 20 30 40 60 nuclei were a priori assigned to either diploid, tetraploid, or octaploid category. Nuclei were selected a t Number of cells random, starting in one corner of the slide and scanFIG.1. The progressive mean with standard deviation of the IOD is ning along the abscissa until the end and backward at shown for a number of diploid nuclei of an imprint. The coefficient of a distance of 2 fields of view. In a slide, 51 (range 48variation remained constant (3.9%)after 25 nuclei. 53) diploid, 50 (range 48-52) tetraploid, and 25 (range 23-28) octaploid nuclei were measured. The automatic segmentation option was used. From on cytometric features such as relative nuclear DNA the digitized nuclear image, the following cytometric content (IOD) and nuclear area and texture features. For this purpose, fine-needle aspirates (FNA), im- features were calculated: IOD (amount of DNA), ploidy prints, and smears of suspensions were made from one value, area, and a number of Markov chain grain dependent texture features. These were contrast, inverse normal human liver specimen. difference moment, difference variance, difference entropy, coefficient of variation (3) and blobness (141, and MATERIALS AND METHODS a statistic feature standard deviation (independent of Specimen Preparation Small pieces of one normal human liver obtained a t grain) (3). The DNA 4ct2c ratio and 8c/2c ratio were autopsy within 8 h after death were snap-frozen. After calculated from the tetraploid and octaploid peaks in defrosting, three different types of cytological samples the DNA histogram by dividing the mean IOU value of were prepared on separate albumin-coated slides (19). the tetraploid and octaploid peak by the mean IOD First, FNA were taken from the blocks. Subsequently, value of the diploid peak. The deviation from linearity imprints were prepared with a maximum of 6 imprints was determined by dividing the difference between the of each side. From the same blocks a suspension was measured and hypothetical value by the hypothetical made by gentle mechanical dissociation of the liver in value. The deviation of linearity was expressed in perPBS buffer at 4 "C. After filtering through a 50-km centages. IOD in arbitrary units (AU) for a nuclear population gauze and centrifugation a t 400 g for 5 min, the pellet was resuspended in a similar volume of PBS and in a specimen was obtained by multiplying the IOD smears were spread out. All specimens were air-dried value by 1.87 and dividing by the mean rat tetraploid for 30 min and then fixed in 4% buffered formaldehyde IOD value. The ratio rat tetraploid DNAhuman dipfor 30 min followed by a short rinse in tap water and loid DNA leads to 1.87. then air drying. Specimens were stored a t room temStatistics perature in a closed box until staining. Mean and SD of the cytometric features were deterStaining mined for the diploid, tetraploid, and octaploid nuclei of All specimens were Feulgen stained according to the each specimen, as well as for the type of tissue samCAS quantitative DNA staining kit (Becton and Dick- pling used. In order to establish the minimum number
487
CELL EXTRACTION METHODS AND CYTOMETRIC FEATURES
of nuclei to be measured in a specimen, the progressive mean was calculated (7,171. In each nuclear population, the Kruskal-Wallis test was used t o compare the difference between the 3 sampling methods for diploid, tetraploid, and octaploid nuclei separately. Each of these comparisons was undertaken a t the 0.01 significance level. If an overall difference was found, the Mann-Whitney U test was used for pairwise comparison of the different sampling methods at the 0.003 level. The null hypothesis was that no differences were present. The significant levels were chosen so that the probability of rejecting one of the null hypothesis for any of the features was a t most 0.09.
RESULTS All specimens showed strongly stained round nuclei with sharp demarcation and a few nuclei with inhomogeneous staining, and irregular edges were present. For measurements, only strong stained round nuclei with sharp demarcation were used. An example of the progressive mean and standard deviation of the progressive mean is shown for the IOD of diploid nuclei in an imprint (see Fig. 1). This is typical of all the slides. The progressive mean and standard deviation of the progressive mean were more or less stable for all specimens after measuring 25-30 nuclei. Therefore, a reliable statistic of DNA ploidy and CV for a single stemline with no or little proliferation is obtained after 30 nuclei are measured. Mean and standard deviation of the cytometric features are shown for diploid, octaploid, and tetraploid nuclei measured in FNA, imprints, and suspensions in Table 1. The results of the Kruskal-Wallis and MannWhitney U tests are shown in Table 2. For the 3 types of specimens, the IOD of diploid nuclei varied slightly. These differences were not significant ( p < 0.01). However, a comparison of imprints and suspensions reveals a significant difference for tetraploid and octaploid nuclei. The average coefficient of variation for IOD from each nuclear population remains below 6% in all specimen types. Nuclear area was lowest for imprints and highest in suspensions for the three nuclear types. For all nuclear populations, a significant difference ( p < 0.003) was present between nuclear areas in the 3 tissue sampling methods. These data prove that the feature nuclear area is very sensitive to specimen preparation. Comparison of the tissue sampling methods reveals significant differences within all nuclear populations for the majority of the texture features. For the 3 types of specimens, the ratios between 4cI2c and 8cI2c peaks and deviation from linearity are shown in Table 3. Note that the 4c/2c ratios are close to the theoretical value of 2 and that the deviation from linearity is approximately 3%. The 99%confidence limits were well within the range of 1.9-2.1, irrespective of the type of specimen measured. The 8cI2c ratios were close to the expected value of 4. The deviations of lin-
Table 1 Mean and Standard Deviation (sd) of Integrated Optical Density (IOD), Nuclear Area, and Texture Features (T) for Diploid, Octaploid, and Tetraploid Nuclei in FNA, Imprints, and Suspensions"
Diploid IOD [AU]
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev Tetraploid IOD rAUl
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev Octaploid IOD [AU]
cv [%I
Area [pm2] T contrast T inv d mom T d variance T d entropy T blobness T coef var T st dev
Imprint mean sd
FNA mean sd
Suspension mean sd
0.95 f 0.06 4.7 2 1.4 42.4 f 5.3 48 t 10 0.34 f 0.02 23.3 t 4.8 0.60 t 0.02 0.31 f 0.02 0.25 t 0.02 0.10 f 0.01
1.0 t 0.06 4.8 f 1.8 56.6 t 7.0 25.8 f 7.3 0.40 f 0.03 13.6 2 3.5 0.60 f 0.02 0.33 2 0.01 0.25 0.01 0.07 2 0.01
0.98 f 0.05 5.3 f 1.7 67.1 t 1.9 17 f 1.7 0.42 f 0.01 8.7 2 0.9 0.57 f 0.01 0.35 0.01 0.26 f 0.01 0.06 0.02
1.88 2 0.08 4.6 f 1.2 70.3 t 9.7 44 f 13 0.38 f 0.02 22 f 6 0.55 5 0.02 0.35 2 0.01 0.22 t 0.01 0.09 f 0.01
2.03 0.12 4.7 t 0.8 96 f 17 26 t 6 0.41 f 0.03 13 f 3 0.56 t 0.04 0.36 f 0.03 0.25 0.03 0.07 2 0.01
1.96 t 0.12 5.8 t 1.7 116 f 5.6 18.6 2 2.1 0.45 2 0.01 10 1 0.52 ? 0.02 0.4 f 0.01 0.28 t 0.01 0.07 t 0.001
3.80 2 0.14 4.4 f 1.5 119 11 55 f 18 0.42 2 0.01 29 f 9 0.51 f 0.01 0.40 f 0.01 0.26 0.01 0.13 t 0.02
4.06 f 0.21 4.3 f 1.1 155 t 25 30 f 7 0.43 t 0.03 16 t 3 0.52 f 0.04 0.39 f 0.03 0.26 0.03 0.15 f 0.21
3.92 k 0.08 5.4 t 1.2 207 t 22 18.3 f 1.8 0.49 f 0.02 11.2 f 1.0 0.47 k 0.02 0.45 f 0.02 0.31 +- 0.02 0.08 +- 0.01
* *
* *
*
*
* *
*
9nv = inverse; d = difference; coef var = coefficient of variation; st dev = standard deviation texture.
earity were slightly higher at this end of the density scale (74.9%). For the 3 types of specimens, the 99% confidence limits were within the 3.7-4.3 region.
DISCUSSION This study deals with the variation in cytometric features and accuracy and linearity of DNA image analysis when measured in imprints, FNA and smears of suspensions. The results of the present study on nuclei of one normal human liver specimen show that differences in tissue sampling methods have little or no influence on DNA measurements (see Table 1). For measurements of normal (diploid) cells from several human organs, Taylor et al. (20) showed a high degree of accuracy and reproducibility of DNA content measurements. Peaks in the DNA histogram with a DNA index between 0.9 and 1.1 were regarded as diploid (20). The results of the Taylor study (20) are in complete concordance with ours.
488
THUNNISSEN E T AL.
Table 2 Summary of the Kruskal-Wallis (K-W) and Mann-Whitney U Test"
K-W FNA-IMP FNA-SUS IMP-SUS Dipl 0.03 * ns Tetr
