Endocrine Abstracts

JOINT1623

Background: Histological analysis of adrenal tissue, particularly through immunohistochemical (IHC) staining and modern RNA-scope techniques, is pivotal in adrenal disease research and diagnosis. Traditionally, IHC evaluation relies on semi-quantitative methods such as the H-score, which offers the advantage of a straightforward, flexible, and numerical scaling system that integrates both the intensity and distribution of staining. However, these manual approaches are time-intensive, subjective, and prone to inter- and intraobserver variability, resulting in limited reproducibility.

Methods and results: To address these challenges, we developed a widely applicable protocol for adrenal tissue analysis based on automated cell detection using QuPath, an open-source digital pathology platform. Our method was evaluated for accuracy, robustness, and reproducibility and will additionally be assessed in direct comparison to the H-score. The protocol includes optimized cell detection strategies tailored to specific adrenal tissue types (normal gland (n=3), adenomas (n=39) and carcinomas (n=240), non-adrenal controls n=40)), adaptations for various unique staining patterns, assessment of tissue microarrays, and practical solutions for scenarios with restricted computational capacity. Furthermore, QuPath provides advanced AI tools to detect the presence of potentially confounding cells, such as blood cells, connective tissue or fat. We optimized these tools for adrenal tissue, offering practical insights to enhance analytical precision. QuPath is also capable of analyzing RNA-scope stainings (n=59 specimens, assessed for three targets respectively), which is traditionally performed semi-quantitatively, whereby a subset of cells and RNA spots are counted to calculate a spots-per-cell ratio. To enhance efficiency and reproducibility, the manufacturer recommends the use of software-based analysis. Utilizing QuPath’s capabilities for automated cell detection and RNA spot identification significantly streamlines the evaluation process, providing a more representative and reliable assessment of RNA expression across entire tissue sections. Finally, we conclude our protocol by showcasing some of QuPath’s visualization features on the adrenal gland, providing the necessary tools to get the most out of staining experiments.

Conclusion: Computational techniques have the potential to fundamentally revolutionize the assessment of adrenal tissue. With automated image analysis, large amounts of data can be evaluated more efficiently, reproducibly, and objectively, especially regarding marker expression. The additional use of artificial intelligence opens up the possibility of specifically recognizing distinct cell types and accounting for interfering tissue.