Title : Artificial intelligence assisted multimodal histological and viability assessment of cartilage tissue preservation
Abstract:
Tissue banks encounter a significant challenge in maintaining viable cartilage for clinical uses such as osteochondral transplantation, trauma repair, and regenerative techniques. Cartilage is a uniquely specialized, avascular tissue where the integrity of its extracellular matrix (ECM) and the survival of chondrocytes are essential for its proper function. Nonetheless, cartilage quickly deteriorates under typical preservation conditions, restricting its shelf life and practical application in clinical settings. The development of efficient preservation techniques, along with robust, impartial tools for evaluating matrix and cell quality, is crucial to enhance transplant outcomes and minimize tissue loss. Conventional histological and viability tests are commonly employed, but they are limited by subjective analysis and variability in image measurement.
This research presents an Artificial Intelligence (AI) enhanced pipeline for the consistent, repeatable, and quantitative assessment of cartilage matrix retention and cellular viability across various preservation solutions. Femoral condyle explants were subjected to incubation in three different preservation media (Solution A: negative control, Solution B: experimental treatment, Solution C: positive control) under two temperature conditions (4°C and 37°C with 5% CO2) and were evaluated at 0, 14, and 28 days. Histological staining with Safranin O and Alcian Blue was used to assess the content of sulfated and non-sulfated glycosaminoglycans (GAG), respectively. To evaluate cell viability, Resazurin assays were used to assess metabolic activity, while Live/Dead staining allowed visualization of membrane integrity. Fluorescence signals were analyzed to quantify viable and non-viable cells over time. Tools for image analysis driven by AI were developed to automatically measure color channel intensities from histological images, reflecting GAG content in a pixel-based and impartial manner. The intensity of the red channel was utilized for Safranin O, while the intensity of the blue channel was employed for Alcian Blue. Intensity values were adjusted to a baseline reference image from day 0, allowing for inter-sample comparison and conversion to relative preservation percentages.
Findings showed that Solution B successfully maintained matrix elements and cell viability at day 14, particularly under incubation at 37°C with 5% CO2, with red channel retention of 91.90%. Live/Dead analysis confirmed higher proportions of viable cells at this point, while Resazurin assays showed sustained metabolic activity. However, both metrics declined significantly by day 28. Alcian Blue staining exhibited a similar trend of decreasing blue intensity. Solution A consistently demonstrated inadequate preservation, whereas Solution C showed moderate effectiveness, confirming its role as a control but not surpassing the experimental formulation.
This research highlights the capability of incorporating AI-driven image analysis into histological and functional evaluation processes. The automated extraction of staining intensities and viability indicators enhanced analytical throughput, eliminated inter-observer bias, and allowed for high-resolution temporal monitoring of tissue quality. These data-driven methods are beneficial in enhancing preservation protocols for cartilage and other ECM-abundant tissues. These results support the role of AI as an auxiliary resource in the preclinical assessment and normalization of tissue engineering approaches.
