The rules of acquiring digital images of the IHC stained tissue samples in terms of colour temperature and tintsf the IHC stained tissue samples in terms of colour temperature and tint 1 – Nalecz Institute of Biocybernetics and Biomedical Engineering PAS, Warsaw, Poland. 2 – Military Institute of Medicine, Warsaw, Poland. 3 – Warsaw University of Technology, Warsaw, Poland. MATERIALS AND METHODS RESULTS CONCLUSIONS Anna Korzynska 1 , Lukasz Roszkowiak 1 , Dorota Pijanowska 1 , Wojciech Kozlowski 2 , Tomasz Markiewicz 2,3 BACKGROUND This study concerns the problem of image unification in a sense of colour and contrast as a part of projected internet platform software for the automatic image analysis. These investigations are mainly focused on tissue samples immunohistochemically stained with 3,3'-Diaminobenzidine and Haematoxylin (DAB&H) to perform the quantification of nuclei. colour variation tissue characteristics sample preparation staining reactivity staining procedures sample thickness digital image acquisition variety of acquisition devices adjusted settings focus white balance aperture exposure time etc. adjusted WB lower light temperature higher light temperature Methods: Images without and with WB correction were compared with the reference image and among themselves to find if the image acquisition process or image manipulations lead to loss of information. Two types of reference images have been used: (1) image with statistical descriptors closest to the averaged for all images acquired with adequate WB settings; (2) artificial image developed as the average of images; all images acquired with adequate WB settings. Each field of view has been acquired 225 times with various microscope light temperature (from 5 up to 12 V) and camera WB settings (from 5 up to 12). Inadequate WB setting causes colour diversity; images acquired with lower WB setting than light temperature are blue tinted, while acquired with higher WB setting are yellow tinted. The images received with the developed internet platform are very different, what makes the development of effective image analysis software difficult. Thus, the image unification should be performed as a part of image preprocessing. In this investigation we focus on the dependence of colour variation in images captured using matched and mismatched camera white balance (WB) setting to microscope's light colour temperature. The mismatches in WB adjustment can be easily handled on the stage of the image acquisition. However, if analysis is done with fully automatic software on images or virtual slides collected and sent by various users, who have used various acquisition tools, images should be unified in pre-processing stage. Our aim is to establish rules of acquiring digital images to avoid loss of important information, taking in to amount segmentation criteria, which is responsible for unreliable results of analysis. image segmentation image in accepted range? 1 Image contrast selective changes in colour tinta chroma colour temperature histogram Manipulations yes no Materials: Methods of comparison Mean Square Error SNR Structural SIMilarity Visual Assessment disease meningioma immunohistochemistry staining DAB&H against Ki67 field of view; no 6 images with empty background field of view acquired; no 255 options of lamp filament 15 options of camera WB settings 15 8000K clear blue sky 1000K candle light 4500K fluorescent light 3000K tungsten microscope’s lightbulb 5500K electronic flash 1.Acquiring under mismatched settings causes the images to be bluish or yellowish and that intensity of discolouration is bigger if the difference in camera settings and microscope’s light temperature is bigger; 2.If the discolouration is reduced by manual or automatic procedures of the WB correction, the contrast in luminance decreases in such a degree, as this difference increases; 3.The best light colour temperature appears when light filaments are powered by the voltage of 7.5 -9.0 V, based on images acquired with matched settings of camera’s WB to the microscope's light temperature; 4.Powering light with less or more voltage increases MSR but decreases SNR and SSIM; 5.Deconvolution leads to relatively stable brown colour fraction in comparison to brown calculated from RGB channels. Methods of image comparison, based on mathematical background, require the reference image and suffer from biased convergence with human judgment, as they are far from human perception system. The manual WB correction has been done using CameraRaw 4.1 adjusting mainly: 1. WB correction (manually shown gray point in clean background or/and stroma); 2. exposure correction (by distribution of extra light simulation), 3. tonal curve manipulation (arbitrary). All manual manipulations has been done regarding the reference image. Left side figure shows the grand means of: (1) brown colour deconvolution – DAB, (2) blue colour deconvolution – Haematoxylin, (3) luminance from Lab colour space, (4) brown colour estimated from RGB channels according to Ruifrok definition; calculated in groups of images acquired with the same WB adjustments showing versions dependence from deepness of mismatching between WB camera settings and microscope’s light temperature. Almost independent appears to be (1) which varies in range of 21 gray levels. But (2) shows highest values in middle region of figure for groups of images acquired with matched or almost matched WB while for greater mismatching (bellow b(n+10)o(n) or over b(n-10)o(n)) the values of grand mean decrease in both directions. It means that contrast between immunopositive (brown – DAB) and immunonegative (blue – Haematxylin) objects is the greatest in images acquired with matched or almost matched WB while this contrast decrease with mismatching increase. The function of the grand mean of (3) decrease from deepest yellow to deepest blue showing that images with the mismatching WB are relatively lighter (yellow end) or darker (blue end). While (4) function of grand mean increases; it is relatively low for big WB mismatch in yellow end and relatively high for opposite blue end. It means that blue colour calculated according to Ruifrok formula is dependent on the blue colour channel RGB, while blue after deconvolution is more similar to G channel in RGB. Right side figure shows the function of aggregated grand mean values of: MSE, SNR and SSIM for 6 images in groups of images acquired with matched WB but under different light conditions. The biggest similarity to reference image, the biggest signal to noise ratio and also smallest mean squared error is visible in the middle of the figure for values between 7,5 and 9. After the visual assessment, the preferred value is 9. DISCUSSION 1.Anna Korzynska, et al.: The influence of the microscope lamp fprocess of digital images of histological ilament colour temperature on the slides acquisition standardization. Diagn Pathol 2014, 9(Suppl 1):S13, In Press. 2.AC Ruifrok, et al.: Quantification of histochemical staining by color deconvolution. Anal Quant Cytol / the International Academy of Cytology [and] American Society of Cytology 23.4 (2001): 291-299. 3.PJ Tadrous: Digital stain separation for histological images. J Microsc 240.2 (2010): 164-172. Acknowledgements This investigation is supported by National Centre for Research and Development of Poland from Applied Research Programme project no PBS2/A9/21/2013 for the years 2013-2016. REFERENCES