Assessment of Computational Histopathology in Thoracic Aortic Aneurysm

Abstract

Background and Hypothesis: Thoracic aortic aneurysm (TAA) histopathology includes elastic fiber (EF) abnormalities, mucoid extracellular matrix (MECM) accumulation, and smooth muscle derangement in the aortic medial layer. While semi-quantitative grading of these characteristics is a standard practice, computational characterization of medial layer components may facilitate novel quantitative analyses at higher throughput. We hypothesized that computational results would correlate with results of semi-quantitative grading of aortic histopathology.

Experimental Design: Formalin-fixed, paraffin-embedded human aortic tissue sections were stained with Movat’s pentachrome to characterize aortic microstructure. Sections were also immunostained for nitrotyrosine residues to assess oxidative stress. Samples were initially graded semi-quantitatively by two independent blinded readers. Next, computational histopathology software was used a) to quantify the proportions of EF, MECM, and cellular area in the medial layer of pentachrome-stained sections and b) to quantify the distribution and intensity of positive nitrotyrosine staining in immunostained sections. Association between semi-quantitative grading and computed values was tested with ANOVA.

Results: The cohort included 74 participants who underwent prophylactic aortic replacement for TAA and 23 healthy controls. The mean age was 54±17 years. On average, EFs accounted for 49% (range 6-90%) of medial tissue area, whereas MECM accounted for 25% (1-73%). The overall semi-quantitative grade of medial degeneration severity was associated with decrease in EF fraction (p=0.02). The grade of EF thinning also strongly correlated with decrease in EF fraction (p=1x10^-6). Meanwhile, grade for accumulation of MECM was associated with increase in MECM (p=0.004). Increased semi-quantitative grading for nitrotyrosine levels was associated with increased nuclear signal optical density (p=9x10^-10) and greater percentage of cells labeled as strongly positive (p=8x10^-10).

Conclusion and Potential Impact: We observed significant correlations between computed quantitative values and semi-quantitative grading. This suggests that computational histopathology is a valid method for investigation of human TAA tissues.