Quantitative evaluation and digitalization of elastic fibers in thoracic aortic aneurysms

Abstract

Introduction: In this study, we initially conducted a quantitative analysis of aortic tissues using conventional histopathologic grading methods. Subsequently, we established a computer-based digitalized semiautonomous image processing pipeline to segment and enhance the principal features of tissue changes, focusing mainly on the qualities of elastic fibers.

Methods: The study collected aortic dilation specimens from 21 patients with a bicuspid aortic valve (mean age, 49,4 ± 14.5 years; aortic diameter, 51.8 ± 7.5 mm), 12 patients with a normal aortic valve (mean age, 64.1 ± 8.6 years; aortic diameter, 53.6 ± 4.0 mm), 3 patients with Marfan syndrome (mean age, 37.1 ± 19.5 years; aortic diameter, 51.5 ± 4.9 mm), and 3 donors (cadaver controls; mean age, 82.0 ± 5.0 years). Tissue samples were fixed in a formalin-buffered solution and prepared for staining with hematoxylin and eosin, Elastica van Gieson, and Alcian blue/van Gieson. Fibrosis, atherosclerosis, medionecrosis, cystic medial necrosis, smooth muscle cell orientation, elastic fragmentation, and inflammation were graded. Histological images were obtained using the conventional method through microscopic observation. Illumination was corrected by low-pass and median filtering, and staining intensity was estimated by k-means clustering in the Commission Internationale de l'Eclairage L*a*b* (CIELAB) color space. Segmented elastic fibers were then corrected by morphological filters to remove artifacts and join dissected fiber segments. Finally, the minimal between-fiber distance was averaged over columns (width, 120px), and the average number of elastic fiber segments was calculated. The accuracy (specificity and sensitivity) of digitalization was calculated among the study groups.

Results: The average number of fibers per area (p = 0.019) and the average distance of fibers per area (p = 0.0015) differed significantly among the study groups. The average grading of the pathological conditions showed no statistical correlation within the groups. There was a significant correlation between the degree of elastic fragmentation observed microscopically and the average number of fibers (p = 0.044) and the average distance between fibers (p = 0.026) analyzed digitally. The elastic fiber fragmentation was seen as the most prominent feature. The area under the curve was 0.698 in the bicuspid aortic valve group, indicating a high number of fibers; therefore, the model can predict whether a bicuspid aortic valve or another valve is present. This result was significant (P < 0.05).

Discussion: Our study compared the conventional method involving microscopic observation and computer-based digitalization. We established a semiautonomous image processing pipeline to segment and enhance the principal features of tissue changes, with a focus mainly on the qualities of elastic fibers. The digitalized method opens new possibilities of analyzing numerous histological slides in a short period. Furthermore, it presents the possibility of performing intraoperative tissue analysis in patients undergoing cardiovascular surgery.

Conclusion: This study demonstrates the potential of combining traditional histopathological techniques with advanced digitalized analysis to improve the accuracy and efficiency of evaluating aortic tissue changes. The digitalized method not only complements conventional methods but also offers a more objective and reproducible assessment, highlighting its applicability in clinical and intraoperative settings.