Quantifying Histochemical Stains Using Whole Slide Imaging
Quantifying Histochemical Stains Using Whole Slide Imaging
Five samples underwent repeated scanning. The H:E ratio for each is shown in Table 3. The ratios for the samples are all under 1, because the amount of eosin in the tissue was greater than the amount of haematoxylin.
The repeatability of the measurement on the same scanner, on the same day, was very high (Pearson correlation coefficient 0.995, p <0.001). The mean magnitude of difference between the two paired measurements was 0.47%; the SD was 2.5E-3 and variance was 22.0E-6). A scatterplot and Bland–Altman plot showed no systematic bias in the measurements.
The reproducibility of the H:E ratio measurement on the same scanner, on different days (Days 0, 12 and 64) was lower, with a mean difference of 8.32%, an SD of 7.6E-3 and variance of 114.8E-6. There was no systematic change in H:E ratio across time (ie, no trend towards increasing or decreasing).
The reproducibility of the measurement on different scanners on the same day was calculated using the second day 64 scans on all four instruments. This showed a strong correlation between the two measurements (Pearson correlation coefficient 0.996–0.999), but higher variability than the other two comparisons with a mean difference of 7.00%, SD of 22.3E-3 and variance of 519.0E-6.
Figure 2 shows a boxplot of the H:E ratios for the three automated staining machines used in the experiment. The median H:E ratio for staining machine 1 was lower than the other two machines (0.69 vs 0.81 and 0.93, p<0.01). SDs for the H:E ratios were 0.11, 0.07 and 0.37 for staining machines 1, 2 and 3, respectively.
(Enlarge Image)
Figure 2.
Boxplot of H:E ratio of three different staining machines. Median H:E ratios were 0.69, 0.81 and 0.93, respectively. Two example images are included to illustrate appendix samples with H:E ratios of approximately 0.69 (bottom) and 0.93 (top).
The granular differences in H&E intensity between the staining instruments are demonstrated in figure 3. Statistically significant differences in the intensity of individual stains were seen—machines 1 and 2 had a higher eosin intensity than 3, and machine 2 had a higher haematoxylin intensity than 2 and 3 (p<0.05). H&E intensity varied independently of each other in most cases.
(Enlarge Image)
Figure 3.
Scatterplot of H&E for three different staining machines. The plot shows the results of experiment 2 where 30 serial sections of appendix were stained on three different machines. Median H&E intensities are plotted on X and Y axes, respectively, and the number at each point indicates which staining machine was used. The plot shows that H&E vary between samples and staining machines. Although some samples show co-variation of H&E, many do not.
Plotting the H:E ratio over time (see figure 4) confirmed that staining machine 3 showed significant variability in staining, with an H:E ratio ranging from 0.7 to over 1.7. The other two staining machines were more consistent, with H:E ratios varying between 0.58 and 0.75. Two large upward spikes in H:E ratio were observed—one with machine 3 at the start of the experiment and one with machine 1 in the middle of the experiment.
(Enlarge Image)
Figure 4.
Line plot of H:E ratio over time for three staining machines. Significant variability in staining is seen, with H:E ratio from 0.7 to 1.7. The large upward spikes seen in the H:E ratio correspond to the end of the maintenance cycle for two machines, at which the stains and reagents were replenished, returning the H:E ratio to acceptable values.
Figure 5 shows the H:E ratio for the six laboratories in the experiment. The mean H:E ratio across the six sites was 0.64, with an SD of 0.12. Five of the laboratories had similar H:E ratios between 0.57 and 0.63, but one laboratory (laboratory E) had a significantly higher H:E ratio of 0.89 (p<0.05).
(Enlarge Image)
Figure 5.
Scatterplot of H:E ratio for six regional laboratories. Most samples have H:E ratios between 0.57 and 0.63, but one hospital laboratory (E) had a higher ratio of 0.89, indicating that samples were significantly more haematoyxlinophilic.
Even between laboratories with a similar H:E ratio, there was significant variability in the intensity of the H&E components of the stains. For example, although laboratories A and D had similar H:E ratios (0.57 and 0.62, respectively), both median H&E intensities were slightly higher by 32% and 21%, respectively, in laboratory D (but these differences were not statistically significant).
Results
Experiment 1: Validation of Stain Quantification Method
Five samples underwent repeated scanning. The H:E ratio for each is shown in Table 3. The ratios for the samples are all under 1, because the amount of eosin in the tissue was greater than the amount of haematoxylin.
The repeatability of the measurement on the same scanner, on the same day, was very high (Pearson correlation coefficient 0.995, p <0.001). The mean magnitude of difference between the two paired measurements was 0.47%; the SD was 2.5E-3 and variance was 22.0E-6). A scatterplot and Bland–Altman plot showed no systematic bias in the measurements.
The reproducibility of the H:E ratio measurement on the same scanner, on different days (Days 0, 12 and 64) was lower, with a mean difference of 8.32%, an SD of 7.6E-3 and variance of 114.8E-6. There was no systematic change in H:E ratio across time (ie, no trend towards increasing or decreasing).
The reproducibility of the measurement on different scanners on the same day was calculated using the second day 64 scans on all four instruments. This showed a strong correlation between the two measurements (Pearson correlation coefficient 0.996–0.999), but higher variability than the other two comparisons with a mean difference of 7.00%, SD of 22.3E-3 and variance of 519.0E-6.
Experiment 2: Intralaboratory Quantification of Stain Intensity
Figure 2 shows a boxplot of the H:E ratios for the three automated staining machines used in the experiment. The median H:E ratio for staining machine 1 was lower than the other two machines (0.69 vs 0.81 and 0.93, p<0.01). SDs for the H:E ratios were 0.11, 0.07 and 0.37 for staining machines 1, 2 and 3, respectively.
(Enlarge Image)
Figure 2.
Boxplot of H:E ratio of three different staining machines. Median H:E ratios were 0.69, 0.81 and 0.93, respectively. Two example images are included to illustrate appendix samples with H:E ratios of approximately 0.69 (bottom) and 0.93 (top).
The granular differences in H&E intensity between the staining instruments are demonstrated in figure 3. Statistically significant differences in the intensity of individual stains were seen—machines 1 and 2 had a higher eosin intensity than 3, and machine 2 had a higher haematoxylin intensity than 2 and 3 (p<0.05). H&E intensity varied independently of each other in most cases.
(Enlarge Image)
Figure 3.
Scatterplot of H&E for three different staining machines. The plot shows the results of experiment 2 where 30 serial sections of appendix were stained on three different machines. Median H&E intensities are plotted on X and Y axes, respectively, and the number at each point indicates which staining machine was used. The plot shows that H&E vary between samples and staining machines. Although some samples show co-variation of H&E, many do not.
Plotting the H:E ratio over time (see figure 4) confirmed that staining machine 3 showed significant variability in staining, with an H:E ratio ranging from 0.7 to over 1.7. The other two staining machines were more consistent, with H:E ratios varying between 0.58 and 0.75. Two large upward spikes in H:E ratio were observed—one with machine 3 at the start of the experiment and one with machine 1 in the middle of the experiment.
(Enlarge Image)
Figure 4.
Line plot of H:E ratio over time for three staining machines. Significant variability in staining is seen, with H:E ratio from 0.7 to 1.7. The large upward spikes seen in the H:E ratio correspond to the end of the maintenance cycle for two machines, at which the stains and reagents were replenished, returning the H:E ratio to acceptable values.
Experiment 3: Interlaboratory Comparison of Stain Intensity
Figure 5 shows the H:E ratio for the six laboratories in the experiment. The mean H:E ratio across the six sites was 0.64, with an SD of 0.12. Five of the laboratories had similar H:E ratios between 0.57 and 0.63, but one laboratory (laboratory E) had a significantly higher H:E ratio of 0.89 (p<0.05).
(Enlarge Image)
Figure 5.
Scatterplot of H:E ratio for six regional laboratories. Most samples have H:E ratios between 0.57 and 0.63, but one hospital laboratory (E) had a higher ratio of 0.89, indicating that samples were significantly more haematoyxlinophilic.
Even between laboratories with a similar H:E ratio, there was significant variability in the intensity of the H&E components of the stains. For example, although laboratories A and D had similar H:E ratios (0.57 and 0.62, respectively), both median H&E intensities were slightly higher by 32% and 21%, respectively, in laboratory D (but these differences were not statistically significant).
Source...