Abstract
Dear Editor,
This letter is in response to the Letter to the Editor by Van Noorden and Chieco that references our article in the Journal of Histochemistry and Cytochemistry from May 2013, among other publications.
We would like to thank the authors of the letter for their comments on the importance of thorough description of the methodology used when generating data from acquired images. Although we agree with their point on having understanding of the significance of proper methodology use and documentation, we believe that the level of description of the methodology, as often referenced to our prior article (Zabka et al. 2011), is correct and up to par with descriptions of similar methods in recent scientific publications. Nevertheless, to ensure that our main focus on the numerical correlation of image data with other molecular and clinical results would not be misinterpreted as shortage of procedural documentation, we would like to review and/or clarify some of the relevant methodology details that were mentioned in the letter to the editor and were not extensively specified or included in our article.
Procedures for immunohistochemical staining were reviewed in the Materials & Methods section, whereas specific details were explained in a prior publication (Zabka et al. 2011) to which the audience was referred. Immunohistochemical staining intensity of positive regions was not quantified but rather utilized to define a minimum positive intensity threshold, which was set to achieve equivalence with manual scoring by the pathologist on a number of high-power slide fields. Thresholding for “dark staining” and “weak staining” was based on a comparison to the “normal cells” in the control group and set to achieve concordance with interpretation by the pathologist. For hematoxylin and eosin stains, segmentation was based on differences in grayscale intensities across image layers, as well as size/shape of the different image/slide elements.
Color deconvolution for separation of hematoxylin and DAB was performed using the software tools referenced in the original article (Definiens Developer XD; Definiens Inc., Munich, Germany), which are based on color separation algorithms originally published by van der Laak et al. (2000). Last, as stated in the original article, whole-slide images were acquired at 20× magnification using a Mirax slide scanner (Carl Zeiss Microimaging; Thornwood, NY) with the standard factory equipment-mounted optics, which includes a 3 CCD color camera with a pixel resolution of 0.23 µm and white illumination light.
In summary, we would like to emphasize that our original aim was to quantify the pathologist evaluation and provide a continuous scale of values for the morphological/cellular changes observed in the digital slides, so that these could be correlated to other study parameters easily, achieving a more refined interpretation of the changes. Quantitation of microscopic images is becoming increasingly important; however, meaningful data must be extracted from the original histological images. In order to accomplish this, precise, consistent methodology is necessary, and we have confidence that such was used in our research in the cited publication. Thus, the authors of the letter were correct in not assuming that “Garrido et al. (2013) have used inadequate image cytometry technology.”