An integrated deep learning and Bayesian computational method-based quantification of angiogenesis in the chick...

Angiogenesis is the process of the formation of new blood vessels from existing vessels, and it sits directly at the intersection of normal physiology and multiple diseases. Normal physiological activities like embryonic development and tissue homoeostasis require precise regulation. Diseased conditions occur when angiogenesis is dysregulated, which drives tumour growth when excessive or leads to diabetic retinopathy-like conditions when insufficient. Researchers are working on therapeutic interventions, and mechanistic studies have made reliable angiogenesis screening essential. Cell-based assays suffer from limited direct physiological relevance, and animal-based models are complex but expensive and raise ethical questions. In this work, we have developed an integrated deep-learning method for the quantitative estimation of angiogenesis from chick embryo-based experimental data. The chick embryo-based yolk sac membrane model (YSM) provides a complex environment for animal models without the cost of other studies. The model developed here offers a direct optical window providing detailed time-based information. The lack of available image analysis tools capable of handling these specific optical characteristics has hindered this assay for a long time. Subsequently, we have designed a pipeline for angiogenesis image analysis tailored to the YSM assay. Here, we have integrated standard image processing (channel separation and CLAHE enhancement) with deep learning and a Bayesian hierarchical Gompertz Gaussian process model for probabilistic vessel length forecasting. This is validated pharmacologically against the VEGF, anti-VEGF antibody, and hydrocortisone treatment groups. Analysis tools tailored for this assay have not been well reported as the YSM assay has a low signal-to-background ratio (red vessels against a yellow background), which has challenged experienced researchers.