Automated extraction of primary cilia-based biomarkers reveals ageing of cells.

Abstract

We present an automated image-analysis methodology for quantitative assessment of primary cilia in cultured human primary fibroblasts. The proposed approach implements a modular processing pipeline combining deep-learning based nuclear segmentation with intensity-driven segmentation of the ciliary axoneme and basal body. These partial segmentations are integrated using a distance-based association strategy, enabling automated reconstruction of individual cilia and subsequent extraction of geometrical features. Performance was evaluated against manually segmented ground truth. While manual annotations showed a systematic underestimation of cilia length, both automated and manual analyses produced consistent population-level trends and identical statistical discrimination across sample groups. Application of the pipeline revealed a progressive reduction in cilia length and ciliation frequency with increasing passage number, demonstrating the sensitivity of the method to subtle morphological changes. The proposed framework enables scalable, reproducible, and objective quantification of primary cilia morphology and provides a computational tool applicable to high-throughput studies of cellular ageing and related phenotypes. All the code related to this work is available through GitHub: https://github.com/reyesaldasoro/Cilia.