Integrating a deep neural network and Transformer architecture for the automatic segmentation and survival prediction in cervical cancer.

Background: Automated tumor segmentation and survival prediction are critical to clinical diagnosis and treatment. This study aimed to develop deep-learning models for automatic tumor segmentation and survival prediction in magnetic resonance imaging (MRI) of cervical cancer (CC) by combining deep neural networks and Transformer architecture. Methods: This study included 406 patients with CC, each with comprehensive clinical information and MRI scans. We randomly divided patients into training, validation, and independent test cohorts in a 6:2:2 ratio. During the model training, we employed two architecture types: one being a hybrid model combining convolutional neural network (CNN) and ransformer (CoTr) and one of pure CNNs. For survival prediction, the hybrid model combined tumor image features extracted by segmentation models with clinical information. The performance of the segmentation models was evaluated using the Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD95). The performance of the survival models was assessed using the concordance index. Results: The CoTr model performed well in both contrast-enhanced T1-weighted (ceT1W) and T2-weighted (T2W) imaging segmentation tasks, with average DSCs of 0.827 and 0.820, respectively, which outperformed other the CNN models such as U-Net (DSC: 0.807 and 0.808), attention U-Net (DSC: 0.814 and 0.811), and V-Net (DSC: 0.805 and 0.807). For survival prediction, the proposed deep-learning model significantly outperformed traditional methods, yielding a concordance index of 0.732. Moreover, it effectively divided patients into low-risk and high-risk groups for disease progression (P<0.001). Conclusions: Combining Transformer architecture with a CNN can improve MRI tumor segmentation, and this deep-learning model excelled in the survival prediction of patients with CC as compared to traditional methods.

Improved night shift schedule related to the mortality of critically ill patients with Corona Virus Disease 2019.

PURPOSE: To determine the relationship between the improved night shift schedule and the mortality of critically ill patients with Corona Virus Disease 2019 (COVID-19). METHODS: According to the time of the implementation of the new night shift schedule, we divided all patients into two groups: initial period group and recent period group. The clinical electronic medical records, nursing records, laboratory findings, and radiological examinations for all patients with laboratory confirmed Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection were reviewed. Cox proportional hazard ratio (HR) models were used to determine the risk factors associated with in hospital death. RESULTS: A total of 75 patients were included in this study. Initial period group includes 45 patients and recent period group includes 30 patients. The difference in mortality between the two groups was significant, 77.8% and 36.7%, respectively. Leukocytosis at admission and admitted to hospital before the new night shift schedule were associated with increased odds of death. CONCLUSIONS: Shift arrangement of medical staff are associated with the mortality of critically ill patients with COVID-19. The new night shift schedule might improve the continuity of treatment, thereby improving the overall quality of medical work and reducing the mortality of critically ill patients.

Frequency-Diverse Bunching Metamaterial Antenna for Coincidence Imaging.

A frequency-diverse bunching metamaterial antenna for coincidence imaging in the Ka band is proposed in this paper. The bunching metamaterial antenna includes a broadband circular array and a frequency-diverse bunching metalens. Firstly, in order to enhance the bunching characteristic, the broadband circular array is designed based on the 60-degree beamwidth design to generate radiation patterns from 32 GHz to 36 GHz. Then, types of metamaterial elements with different transmission phases are selected to form the frequency-diverse bunching metalens based on a random distribution design and gradient zoom coefficient design. Moreover, the bunching metamaterial antenna is constituted by loading the frequency-diverse bunching metalens to the broadband circular array, which can generate frequency-diverse bunching random radiation patterns with beamwidth less than 100 degrees from 32 GHz to 36 GHz. Furthermore, the performances of the bunching metamaterial antenna, including the reflection coefficient, the radiation efficiency, and the correlation coefficients of radiation patterns at different frequencies are evaluated. Finally, the coincidence imaging experiment is implemented using the bunching metamaterial antenna and the image of the target is reconstructed successfully. The design is verified by simulations and measurements.

Multi-polarization, polarization-independent, wide-angle RCS reduction metasurface based on random phase gradients.

A novel metasurface based on random phase gradients is proposed for radar cross-section (RCS) reduction. In this work, wideband, polarization-independent, wide-angle RCS reduction is realized for both circularly polarized (CP) waves and linearly polarized (LP) waves, respectively. Thus, true polarization-independent RCS reduction is realized by the proposed metasurface. This proposed metasurface is composed of different types of units, and these units do not have certain periods. Under both CP incidence and LP incidence, random phase gradients can be formed on the proposed metasurface. The incidence can be diffused because of these random phase gradients, resulting in multi-polarization, polarization-independent, wide-angle RCS reduction. The 10 dB RCS reduction ranges from 12.6 GHz to 17.0 GHz and 18 GHz to 22 GHz for right-hand circularly polarized incident waves, and from 12.4 GHz to 17.0 GHz and 18.0 GHz to 21.8 GHz for left-hand circularly polarized incident waves. Meanwhile, the 10 dB RCS reduction ranges from 12.0 GHz to 17.0 GHz for x-polarized incident waves and from 13.0 GHz to 17.0 GHz and 17.6 to 21.8 GHz for y-polarized incident waves. Both the simulation and experimental results verify the value of this proposed metasurface in stealth technology.

Self-Adaption Matched Filter and Bi-Directional Difference Method for Moving Target Detection.

In this paper, a self-adaption matched filter (SMF) and bi-directional difference techniques are proposed to detect a small moving target in urban environments. Firstly, the SMF technique is proposed to improve the signal-to-interference-noise ratio (SINR) by using the power factor. The properties of the transmitting signal, the target echoes and the interference and noise are considered during the power factor generation. The amplitude coherent accumulation technique that extracts the coherent amplitude information of echoes after being processed by the SMF, is used to improve the SINR based on multiple measurements. Finally, the bi-directional difference technique is proposed to distinguish the target echoes and the interference/noise. Simulations and experiments are conducted to validate and demonstrate that small moving targets can be detected with high probability using the proposed method in urban environments, even with just one measurement.

Ghost Imaging Based on Deep Learning.

Even though ghost imaging (GI), an unconventional imaging method, has received increased attention by researchers during the last decades, imaging speed is still not satisfactory. Once the data-acquisition method and the system parameters are determined, only the processing method has the potential to accelerate image-processing significantly. However, both the basic correlation method and the compressed sensing algorithm, which are often used for ghost imaging, have their own problems. To overcome these challenges, a novel deep learning ghost imaging method is proposed in this paper. We modified the convolutional neural network that is commonly used in deep learning to fit the characteristics of ghost imaging. This modified network can be referred to as ghost imaging convolutional neural network. Our simulations and experiments confirm that, using this new method, a target image can be obtained faster and more accurate at low sampling rate compared with conventional GI method.