Abstract Tu0064: Vessel tech+: a high-accuracy pipeline for comprehensive mouse retinal vasculature characterization and filopodia identification

Rationale: Mouse retinal vasculature is a well-recognized and commonly used animal model for angiogenesis and microvascular remodeling. Morphological features extracted from the retinal vasculature reflect the vessel’s biological functions and are critical in understanding the physiological and pathological processes of vascular development and disease. However, most feature extractions are still done manually, especially the filopodia. There is an urgent need for an accurate and comprehensive toolbox to automatically characterize morphological changes and analyze filopodia in mouse retinal vasculature. Objective: Develop a comprehensive and highly accurate pipeline to analyze morphological changes and filopodia in the mouse retinal vasculature. Methods and Results: Retinas from postnatal mice at P2-P7 are harvested and retinal vasculatures are visualized by isolectin B4 whole mount staining. A comprehensive software, Vessel Tech, is developed to perform vessel segmentation, network reconstruction and analysis. Vessel Tech relies on recent advancements in the convolutional neural network and techniques from network data analysis to provide high-accuracy vascular identification and comprehensive vascular morphological feature extraction. Compared with existing blood vessel analysis software, Vessel Tech shows great improvement in accuracy. Based on the reconstructed vessel network, various metrics of angiogenesis and micro-remodeling are extracted and statistically analyzed. Furthermore, we upgraded this pipeline, named VesselTech+, to segment the blood vessels and filopodia accurately. Instead of just focusing on the modeling part, we have also experimented with the imaging methods. We varied the image resolution and staining methods to obtain the best data to train our deep neural networks to extract vessel networks and filopodia accurately. Conclusions: We generated and verified a novel pipeline named “Vessel Tech+,” which could automatically process retinal vascular images, reconstruct vessel networks with high accuracy, and assay global and local vascular characteristics. The development of Vessel Tech+ provides a powerful tool to precisely study the physiological and pathological variations during retinal vascular development and vascular diseases.