In this study, we present an automated approach to classify prostate cancer (PCa) whole slide images (WSIs) as high or low cancer aggressiveness using features derived from persistent homology, a tool of topological data analysis (TDA). This extends previous work on the use of these features for representing the characteristics of prostate cancer architecture in region of interest (ROI) images, and demonstrates the value of features derived from persistent homology to predict cancer aggressiveness of WSIs on an ROI basis. We compute persistence on ROI images and summarize persistence as a persistence image. Using this summary we construct a random forest classifier to predict cancer aggressiveness. We demonstrate the potential of persistent homology to capture the architectural differences between low and high grade prostate cancers in a feature representation that lends itself well to machine learning approaches.
Prostate cancer comprises the second most common cancer in men. One of the most powerful and established prognostic indicators of adenocarcinoma of the prostate is the Gleason score, a subjective assessment of the pattern of tumor growth and extent of glandular differentiation in H&E stained histology slides. Despite being the most dominant prostate grading method in use, the Gleason score suffers from high variability between grading pathologists, and due to its 2D nature, fails to effectively capture potentially prognostic information contained in 3D glandular growth patterns. We have previously demonstrated that persistent homology, a subset of topological data analysis (TDA), is effective in generating a quantitative morphological descriptor capable of differentiating Gleason grade in 2D. By capturing glands as loops in 2D, and voids in 3D, persistent homology lends itself naturally to the assessment of 3D glandular growth patterns while maintaining a correspondence to their 2D analogue. Dual-view inverted selective plane illumination microscopy (diSPIM) with a fluorescent H&E analogue was leveraged for volumetric imaging of optically-cleared prostate biopsies. The two orthogonal views of the diSPIM system yielded isotropic resolution in all dimensions, facilitating reconstruction of tissue histology in 3D for quantitative morphological assessment by persistent homology. The use of a nuclei specific hematoxylin analog (DRAQ5), in addition to the isotropic resolution of the system, enabled accurate 3D nuclei segmentation, thereby facilitating application of persistent homology to the corresponding nuclei 3D point clouds. Through TDA a quantitative, reproducible descriptor for 3D prostate cancer morphology will be demonstrated.