Hyperspectral images (HSIs) data that is typically presented in 3-D format offers an opportunity for 3-D networks to extract spectral and spatial features simultaneously. In this paper, we propose a novel end-to-end 3-D dense convolutional network with spectral-wise attention mechanism (MSDN-SA) for HSI classification. The proposed MSDN-SA exploits 3-D dilated convolutions to simultaneously capture the spectral and spatial features at different scales, and densely connects all 3-D feature maps with each other. In addition, a spectral-wise attention mechanism is introduced to enhance the distinguishability of spectral features, which improves the classification performance of the trained models. Experimental results on three HSI datasets demonstrate that our MSDN-SA achieves competitive performance for HSI classification.

BACKGROUND Sedation with propofol injections is associated with a risk of addiction, but remimazolam benzenesulfonate is a comparable anesthetic with a short elimination half-life and independence from cell P450 enzyme metabolism. Compared to remimazolam, remimazolam benzenesulfonate has a faster effect, is more quickly metabolized, produces inactive metabolites and has weak drug interactions. Thus, remimazolam benzenesulfonate has good effectiveness and safety for diagnostic and operational sedation. AIM To investigate the clinical value of remimazolam benzenesulfonate in cardiac surgery patients under general anesthesia. METHODS A total of 80 patients who underwent surgery in the Department of Cardiothoracic Surgery from August 2020 to April 2021 were included in the study. Using a random number table, patients were divided into two anesthesia induction groups of 40 patients each: remimazolam (0.3 mg/kg remimazolam benzenesulfonate) and propofol (1.5 mg/kg propofol). Hemodynamic parameters, inflammatory stress response indices, respiratory function indices, perioperative indices and adverse reactions in the two groups were monitored over time for comparison. RESULTS At pre-anesthesia induction, the remimazolam and propofol groups did not differ regarding heart rate, mean arterial pressure, cardiac index or volume per wave index. After endotracheal intubation and when the sternum was cut off, mean arterial pressure and volume per wave index were significantly higher in the remimazolam group than in the propofol group ( P < 0.05). After endotracheal intubation, the oxygenation index and the respiratory index did not differ between the groups. After endotracheal intubation and when the sternum was cut off, the oxygenation index values were significantly higher in the remimazolam group than in the propofol group ( P < 0.05). Serum interleukin-6 and tumor necrosis factor-α levels 12 h after surgery were significantly higher than before surgery in both groups ( P < 0.05). The observation indices were re-examined 2 h after surgery, and the epinephrine, cortisol and blood glucose levels were significantly higher in the remimazolam group than in the propofol group ( P < 0.05). The recovery and extubation times were significantly lower in the remimazolam group than in the propofol group ( P < 0.05); there were significantly fewer adverse reactions in the remimazolam group (10.00%) than in the propofol group (30.00%; P < 0.05). CONCLUSION Compared with propofol, remimazolam benzenesulfonate benefited cardiac surgery patients under general anesthesia by reducing hemodynamic fluctuations. Remimazolam benzenesulfonate influenced the surgical stress response and respiratory function, thereby reducing anesthesia-related adverse reactions...

No abstract

This paper studies the classification problem of hyperspectral image (HSI). Inspired by the great success of deep neural networks in Artificial Intelligence (AI), researchers have proposed different deep learning based algorithms to improve the performance of hyperspectral classification. However, deep learning based algorithms always require a large-scale annotated dataset to provide sufficient training. To address this problem, we propose a semi-supervised deep learning framework based on the residual networks (ResNets), which use very limited labeled data supplemented by abundant unlabeled data. The core of our framework is a novel dual-strategy sample selection co-training algorithm, which can successfully guide ResNets to learn from the unlabeled data by making full use of the complementary cues of the spectral and spatial features in HSI classification. Experiments on the benchmark HSI dataset and real HSI dataset demonstrate that, with a small number of training data, our approach achieves competitive performance with maximum improvement of 41% (compare with traditional convolutional neural network (CNN) with 5 initial training samples per class on Indian Pines dataset) for HSI classification as compared with the results from those state-of-the-art supervised and semi-supervised methods.