The presence of both endothelial cells (ECs) and mural cells are central to the proper function of blood vessels in health and pathological changes in diseases including diabetes. Although iPSCs-derived vascular organoids (VOs) provide an appealing in vitro disease model and platform for drug screening, whether these organoids recapitulate human disease remains debatable. Here, we show human diabetic (DB)-VOs represent impaired vascular function including enhanced ROS activity, with higher mitochondrial content and activity, increased pro-inflammatory cytokines, and less regenerative potential in vivo. Using single-cell RNA sequencing, we identify all specialized types of vascular cells (artery, capillary, vein, lymphatic and tip cells, as well as pericytes and vSMCs) within vascular organoids, while demonstrating the dichotomy landscape of ECs and mural cells. Furthermore, we reveal basal heterogeneity within vascular organoids and demonstrate differences between diabetic and non-diabetic VOs. Of note, a subpopulation of ECs significantly enrich for ROS and oxidative phosphorylation hallmarks in DB-VOs, may represent early signs of aberrant angiogenesis in diabetes. This study helps to identify key biomarkers for diabetic disease progression and find signalling molecules amenable to drug intervention.
two trained pathologists into 5 classes (IT, XA, PIT, FCA, TCFA). As this led to an unbalanced dataset, we performed data augmentation. Pre-processing, speckle noise reduction (ELF, BM3D) and segmentation (K-means, Gaussian mixture models) were applied in a variety of combinations to evaluate their contribution to classification by five different pretrained ultradeep convolutional neural networks . Saliency maps were implemented to visualize the prediction of the classification. ResultsThe accuracy of ResNet-50 was poor, 40%, but VGG-16, VGG-19, Inception-V3, and DenseNet-121 performed well, achieving 69%, 66%, 78%, and 80% prediction accuracy. Pre-processing appeared essential to obtaining a high accuracy. Further fine tuning of DenseNet-121 led to a training and testing accuracies of 95% and 92%. Saliency maps indicated that DenseNet-121 automatically identified diseased, atherosclerotic plaques, even when they only partially affected the vessel wall. Conclusions Ultradeep learning classifiers when trained with histological data are capable of robust classification of vulnerable plaques even when the plaques are partially affecting the cross section of the vessel wall. This is the first, fully automatic, accurate OCT based vulnerable plaque classifier.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.