No abstract

Imaging of small animals has played an indispensable role in preclinical research by providing high dimensional physiological, pathological, and phenotypic insights with clinical relevance. Yet pure optical imaging suffers from either shallow penetration (up to ~1–2 mm) or a poor depth-to-resolution ratio (~1/3), and non-optical techniques for whole-body imaging of small animals lack either spatiotemporal resolution or functional contrast. Here, we demonstrate that standalone single-impulse photoacoustic computed tomography (SIP-PACT) mitigates these limitations by combining high spatiotemporal resolution (125-µm in-plane resolution, 50 µs / frame data acquisition and 50-Hz frame rate), deep penetration (48-mm cross-sectional width in vivo), anatomical, dynamical and functional contrasts, and full-view fidelity. By using SIP-PACT, we imaged in vivo whole-body dynamics of small animals in real time and obtained clear sub-organ anatomical and functional details. We tracked unlabeled circulating melanoma cells and imaged the vasculature and functional connectivity of whole rat brains. SIP-PACT holds great potential for both pre-clinical imaging and clinical translation.

In this paper, we propose to incorporate convolutional neural networks with a multi-context attention mechanism into an end-to-end framework for human pose estimation. We adopt stacked hourglass networks to generate attention maps from features at multiple resolutions with various semantics. The Conditional Random Field (CRF) is utilized to model the correlations among neighboring regions in the attention map. We further combine the holistic attention model, which focuses on the global consistency of the full human body, and the body part attention model, which focuses on the detailed description for different body parts. Hence our model has the ability to focus on different granularity from local salient regions to global semanticconsistent spaces. Additionally, we design novel Hourglass Residual Units (HRUs) to increase the receptive field of the network. These units are extensions of residual units with a side branch incorporating filters with larger receptive fields, hence features with various scales are learned and combined within the HRUs. The effectiveness of the proposed multi-context attention mechanism and the hourglass residual units is evaluated on two widely used human pose estimation benchmarks. Our approach outperforms all existing methods on both benchmarks over all the body parts.