Deep convolutional neural networks for segmenting 3D in vivo multiphoton images of vasculature in Alzheimer disease mouse models

Abstract

The health and function of tissue rely on its vasculature network to provide reliable blood perfusion. Volumetric imaging approaches, such as multiphoton microscopy, are able to generate detailed 3D images of blood vessels that could contribute to our understanding of the role of vascular structure in normal physiology and in disease mechanisms. The segmentation of vessels, a core image analysis problem, is a bottleneck that has prevented the systematic comparison of 3D vascular architecture across experimental populations. We explored the use of convolutional neural networks to segment 3D vessels within volumetric in vivo images acquired by multiphoton microscopy. We evaluated different network architectures and machine learning techniques in the context of this segmentation problem. We show that our optimized convolutional neural network architecture with a customized loss function, which we call DeepVess, yielded a segmentation accuracy that was better than state-of-the-art methods, while also being orders of magnitude faster than the manual annotation. To explore the effects of aging and Alzheimer’s disease on capillaries, we applied DeepVess to 3D images of cortical blood vessels in young and old mouse models of Alzheimer’s disease and wild type littermates. We found little difference in the distribution of capillary diameter or tortuosity between these groups, but did note a decrease in the number of longer capillary segments (>75μm) in aged animals as compared to young, in both wild type and Alzheimer’s disease mouse models.

Unsupervised Microvascular Image Segmentation Using an Active Contours Mimicking Neural Network

“…In the images domain of two-photons microscopy, Cicek et al [6] proposed a 3D-Unet for vascular segmentation, Teikari et al [32] introduced VesselNN, which is a 2D-3D network architecture for 3D segmentation. Haft-Javaherian et al [13] recently proposed DeepVess, which stands as the current state-of-theart. Additional algorithms for automated volumetric segmentation of microvasculature have been recently described in [9,7,33].…”

“…In all of our transductive learning experiments, we use the same learning rate for both training and fine-tuning. For a fair comparison with supervised methods, we perform the additional training for the same length of time, as it takes to evaluate the test set with the relatively fast DeepVess method [13].…”

High fat diet worsens pathology and impairment in an Alzheimer’s mouse model, but not by synergistically decreasing cerebral blood flow

“…First the raw image stacks of the vasculature were masked to remove the larger surface vessels and regions where the signal to noise was poor using ImageJ (NIH). Masked images were then preprocessed to normalize the image intensity distribution and spatial resolution [42] and to remove motion artifacts [43].…”

“…2. The vasculature network in each image stack was then segmented into binary images using a deep convolutional neural network, DeepVess, we recently developed [42].…”

“…3. Individual capillaries, defined as a segment between two bifurcations, were identified by extracting the vasculature centerlines using sequential dilation and thinning operations [42], in addition to some centerline artifact removal steps [42]. 4.…”