3-D Nuclear chromatin texture analysis using confocal laser scanning microscopy

Abstract

The transformation of a normal cell into a malignant cell is associated with genetic alterations that often result in abnormal chromosome sets (“aneuploidy”) and changes in the distribution of chromatin inside the nucleus. These changes are often subtle and are mostly referred to as “malignancy associated changes” (MACs). Unfortunately, these changes can not easily be detected through the microscope and can also be found in morphologically benign cells. To correctly discriminate between normal and (potentially) malignant cells these changes are therefore best not merely visually assessed but also mathematically quantified by image cytometry as “texture features”. The aim of the studies described in this thesis was to develop methods for quantification of the nuclear chromatin architecture by means of analysis of 3-D texture features as a potential aid in future tissue diagnosis and prognosis assessment of prostate cancer. TO-PRO-3 (a stoichiometric dye) for staining of the tissue specimens was used. After applying the staining procedure, 3-D image stacks were acquired with a confocal microscope. To obtain ploidy and texture feature measurements for single nuclei, a segmentation procedure was applied on the images. Thirty-five features thoughtfully chosen from 4 categories of (3-D) texture features (discrete texture features, Markovian features, fractal features, grey value distribution features) were selected and tested for invariance properties (rotation and scaling) using artificial images. In a pilot study we used the 3-D texture feature analyses to discriminate between benign and malignant prostate nuclei. For each patient, a pathologist selected benign regions and malignant regions, and from those two regions at most 300 nuclei were segmented. Together with the texture feature analysis, ploidy measurements were performed on the segmented nuclei from the 3-D image stack. Finally, we have applied our methods to study whether the differences in mortality rate between Afro-American men and Caucasian-American men having prostate cancer is simply explained by socioeconomic factors or by genetics factors resulting in morphological changes in the nuclear chromatin distribution as well. The best results to discriminate between benign and malignant cell nuclei were obtained when multivariate statistics using Linear Discriminant Analysis was employed instead of ROC analysis. We have shown that we are able to successfully discriminate between benign and malignant nuclei in 89% of the cases. In the successive studies described in this thesis, the many technical difficulties to obtain clinically useful analysis of 3-D nuclear chromatin distributions in prostate tissue have been overcome. It is now possible to successfully perform such analysis, although expensive equipment is required and throughput is low. The 3-D texture features as described in this thesis might well be useful for other types of tissues where there are variations in the distribution of nuclear chromatin. Therefore, by incorporating the sensitivity of nuclear texture features to detect small nuclear chromatin differences an earlier diagnosis can possibly be made.