No abstract

Background: Uneven body-surface thermal distribution is a manifestation of hypoperfusion and can be quantified by infrared thermography. Our aim was to investigate whether body-surface thermal inhomogeneity could accurately evaluate the severity of patients at risk of hypoperfusion. Methods: This was a prospective cohort study in which infrared thermography images were taken from unilateral legs of critically ill patients at high risk of hypoperfusion in a cardiac surgical intensive care unit. For each patient, five body-surface thermal inhomogeneity parameters, including standard deviation (SD), kurtosis, skewness, entropy, and low-temperature area rate (LTAR), were calculated. Demographic, clinical, and thermal characteristics of deceased and living patients were compared. The risk of mortality and capillary refill time (CRT) were chosen as the primary outcome and benchmarking parameter for hypoperfusion, respectively. The area under the receiver operating characteristic curve (AUROC) was used to evaluate predictive accuracy. Results: Three hundred seventy-three patients were included, and 55 (14.7%) died during hospital stay. Of inhomogeneity parameters, SD (0.738) and LTAR (0.768) had similar AUROC to CRT (0.757) for assessing mortality risk. Besides, there was a tendency for LTAR (1%-3%-7%) and SD (0.81°C-0.88°C-0.94°C) to increase in normotensive, hypotensive, and shock patients. These thermal parameters are associated with CRT, lactate, and blood pressure. The AUROC of a combined prediction incorporating three thermal inhomogeneity parameters (SD, kurtosis, and entropy) was considerably higher at 0.866. Conclusions: Body-surface thermal inhomogeneity provided a noninvasive and accurate assessment of the severity of critically ill patients at high risk of hypoperfusion.

No abstract

Despite encouraging results have been achieved in human pose estimation in recent years, it remains challenging problems. The performance may degrade dramatically when the background is similar to the human body parts, and there are very small persons with low-resolution in the image. This paper addresses problems in background-inference and small-person images human pose estimation. To achieve this, a novel pose estimation algorithm is proposed on the basis of person semantic segmentation deep neural network. Different from most previous methods with a single pose estimation model, we generate mixture models with pose estimation and semantic segmentation. We introduce novel generative adversarial model and auxiliary model to realize the semantic segmentation network, which can handle the confusion of the similar regions in the background. In addition, to address the problem of the scale differences between big and small persons' keypoints, we add additional position and channel attention modules to the first two stages of OpenPose. We conduct extensive experiments on COCO and VOC datasets, and compare the proposed method with the most popular state-of-the-art human pose estimation and semantic segmentation frameworks, including MultiPoseNet, Deterton2 and DeepLab V3. Our experimental results show that the proposed method is more accurate than the state-of-the-art algorithms and performs effectively in tackling the complex situations.

No abstract