Color Correction of Stained Tissue Section Images by Histogram Transfer According to Control Images

To present a color correction method by histogram transfer depending upon control tissue image (CTI) differences, and subsequently evaluate its performance.

Images from colon and placenta sections stained by anti-CD34 were used as CTI and/ or sample tissue images (STIs). In total, 36 slides were stained: 1 according to standard procedure and 35 with some variation in the durations or dilutions used for the staining process. For hematoxylin and eosin (20 slides) and Van Gieson (20 slides) stains, colonic mucosa and liver tissues were used. Digital images without normalization were taken by a CCD camera connected to a light microscope and stored on a computer. A software tool was developed in order to find the histogram difference between 2 CTIs and transfer the difference to the STI for achieving a corrected STI (corSTI). sSTI (1 image) and STI and corSTI (for each image) were semiquantitatively scored by 2 observers in blind fashion, and the STI and corSTI scores were compared with the sSTI score. Total optic density (TOD) and median optic density (MOD) and intensity were also calculated by the software.

The STI semiquantitative score was equal to the sSTI in 23.5% of the image; this improved to 76.35% when the corSTI was compared to the sSTI. The concordance of TOD and intensity values of CD34-stained placenta images, as well as TOD and MOD values of H&E-stained colonic mucosa images, with the values calculated for the sSTI, increased following image correction.

These results suggest that histogram transfer depending upon CTIs may be a valuable tool for color correction of tissue section images.