End-to-End Network Intrusion Detection Based on Contrastive Learning

The network intrusion detection system (NIDS) plays a crucial role as a security measure in addressing the increasing number of network threats. The majority of current research relies on feature-ready datasets that heavily depend on feature engineering. Conversely, the increasing complexity of network traffic and the ongoing evolution of attack techniques lead to a diminishing distinction between benign and malicious network behaviors. In this paper, we propose a novel end-to-end intrusion detection framework based on a contrastive learning approach. We design a hierarchical Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) model to facilitate the automated extraction of spatiotemporal features from raw traffic data. The integration of contrastive learning amplifies the distinction between benign and malicious network traffic in the representation space. The proposed method exhibits enhanced detection capabilities for unknown attacks in comparison to the approaches trained using the cross-entropy loss function. Experiments are carried out on the public datasets CIC-IDS2017 and CSE-CIC-IDS2018, demonstrating that our method can attain a detection accuracy of 99.9% for known attacks, thus achieving state-of-the-art performance. For unknown attacks, a weighted recall rate of 95% can be achieved.

Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems

Autonomous driving navigation relies on diverse approaches, each with advantages and limitations depending on various factors. For HD maps, modular systems excel, while end-to-end methods dominate mapless scenarios. However, few leverage the strengths of both. This paper innovates by proposing a hybrid architecture that seamlessly integrates modular perception and control modules with data-driven path planning. This innovative design leverages the strengths of both approaches, enabling a clear understanding and debugging of individual components while simultaneously harnessing the learning power of end-to-end approaches. Our proposed architecture achieved first and second place in the 2023 CARLA Autonomous Driving Challenge's SENSORS and MAP tracks, respectively. These results demonstrate the architecture's effectiveness in both map-based and mapless navigation. We achieved a driving score of 41.56 and the highest route completion of 86.03 in the MAP track of the CARLA Challenge leaderboard 1, and driving scores of 35.36 and 1.23 in the CARLA Challenge SENSOR track with route completions of 85.01 and 9.55, for, respectively, leaderboard 1 and 2. The results of leaderboard 2 raised the hybrid architecture to the first position, winning the edition of the 2023 CARLA Autonomous Driving Competition.

IMH-Net: a convolutional neural network for end-to-end EEG motor imagery classification

As the main component of Brain-computer interface (BCI) technology, the classification algorithm based on EEG has developed rapidly. The previous algorithms were often based on subject-dependent settings, resulting in BCI needing to be calibrated for new users. In this work, we propose IMH-Net, an end-to-end subject-independent model. The model first uses Inception blocks extracts the frequency domain features of the data, then further compresses the feature vectors to extract the spatial domain features, and finally learns the global information and classification through Multi-Head Attention mechanism. On the OpenBMI dataset, IMH-Net obtained 73.90 ± 13.10% accuracy and 73.09 ± 14.99% F1-score in subject-independent manner, which improved the accuracy by 1.96% compared with the comparison model. On the BCI competition IV dataset 2a, this model also achieved the highest accuracy and F1-score in subject-dependent manner. The IMH-Net model we proposed can improve the accuracy of subject-independent Motor Imagery (MI), and the robustness of the algorithm is high, which has strong practical value in the field of BCI.

Adversarially Learning Occlusions by Backpropagation for Face Recognition

With the accomplishment of deep neural networks, face recognition methods have achieved great success in research and are now being applied at a human level. However, existing face recognition models fail to achieve state-of-the-art performance in recognizing occluded face images, which are common scenarios captured in the real world. One of the potential reasons for this is the lack of large-scale training datasets, requiring labour-intensive and costly labelling of the occlusions. To resolve these issues, we propose an Adversarially Learning Occlusions by Backpropagation (ALOB) model, a simple yet powerful double-network framework used to mitigate manual labelling by contrastively learning the corrupted features against personal identity labels, thereby maximizing the loss. To investigate the performance of the proposed method, we compared our model to the existing state-of-the-art methods, which function under the supervision of occlusion learning, in various experiments. Extensive experimentation on LFW, AR, MFR2, and other synthetic masked or occluded datasets confirmed the effectiveness of the proposed model in occluded face recognition by sustaining better results in terms of masked face recognition and general face recognition. For the AR datasets, the ALOB model outperformed other advanced methods by obtaining a 100% recognition rate for images with sunglasses (protocols 1 and 2). We also achieved the highest accuracies of 94.87%, 92.05%, 78.93%, and 71.57% TAR@FAR = 1 × 10-3 in LFW-OCC-2.0 and LFW-OCC-3.0, respectively. Furthermore, the proposed method generalizes well in terms of FR and MFR, yielding superior results in three datasets, LFW, LFW-Masked, and MFR2, and producing accuracies of 98.77%, 97.62%, and 93.76%, respectively.

A Multi-Task Fusion Strategy-Based Decision-Making and Planning Method for Autonomous Driving Vehicles

The autonomous driving technology based on deep reinforcement learning (DRL) has been confirmed as one of the most cutting-edge research fields worldwide. The agent is enabled to achieve the goal of making independent decisions by interacting with the environment and learning driving strategies based on the feedback from the environment. This technology has been widely used in end-to-end driving tasks. However, this field faces several challenges. First, developing real vehicles is expensive, time-consuming, and risky. To further expedite the testing, verification, and iteration of end-to-end deep reinforcement learning algorithms, a joint simulation development and validation platform was designed and implemented in this study based on VTD-CarSim and the Tensorflow deep learning framework, and research work was conducted based on this platform. Second, sparse reward signals can cause problems (e.g., a low-sample learning rate). It is imperative for the agent to be capable of navigating in an unfamiliar environment and driving safely under a wide variety of weather or lighting conditions. To address the problem of poor generalization ability of the agent to unknown scenarios, a deep deterministic policy gradient (DDPG) decision-making and planning method was proposed in this study in accordance with a multi-task fusion strategy. The main task based on DRL decision-making planning and the auxiliary task based on image semantic segmentation were cross-fused, and part of the network was shared with the main task to reduce the possibility of model overfitting and improve the generalization ability. As indicated by the experimental results, first, the joint simulation development and validation platform built in this study exhibited prominent versatility. Users were enabled to easily substitute any default module with customized algorithms and verify the effectiveness of new functions in enhancing overall performance using other default modules of the platform. Second, the deep reinforcement learning strategy based on multi-task fusion proposed in this study was competitive. Its performance was better than other DRL algorithms in certain tasks, which improved the generalization ability of the vehicle decision-making planning algorithm.

The comparison of anastomosis strength and leakage between double-layer full-thickness and single-layer extramucosal intestine anastomosis

Various intestine anastomosis techniques have been studied and used, but which is best is still debated. In our center, double-layer full-thickness intestine anastomosis was still considered as standard. However, a single-layer extramucosal intestine anastomosis has shown favorable results. This study created an anastomotic model to compare the anastomosis strength and leakage between double-layer full-thickness and single-layer extramucosal intestine anastomosis.

METHODS

This experimental study was performed in 20 randomized healthy male pigs, to be included either in Group A (Single-layer extramucosal intestine anastomosis) or Group B (Double-layer full-thickness intestine anastomosis). Enterotomy followed by an end-to-end anastomosis suture was performed in the jejunum. Fourteen days after the operation, any anastomosis leakage and its location was documented. The anastomosis strength was evaluated using manometry. Data were compared between groups using the Mann-Whitney U and Fischer Exact test, considering a significance level of P<0.05.

RESULTS

The overall mean intraluminal anastomotic bursting pressure was 4,257±1,185. Group A had a higher intraluminal anastomotic bursting pressure but was not statistically significant compared to group B (4.726±0.952 vs. 3.787±1.252 kilopascals, P=0.063). One leakage (5%, antimesenteric area) occurred in Group A and three leakages (15%, antimesenteric and mesenteric area) occurred in Group B. However, statistical analysis with Fischer exact showed no significant difference of leakage rate between those groups (P=0.291).

CONCLUSIONS

The anastomosis strength and leakage did not differ significantly between the single-layer extramucosal intestine anastomosis group and the double-layer full-thickness anastomosis group. However, the location of leakage was most common in the antimesenteric area in the double-layer full-thickness anastomosis group.

An ex vivo preliminary investigation into the impact of parameters on tissue welding strength in small intestine mucosa-mucosa end-to-end anastomosis

Background: Tissue welding is an electrosurgical technique that can fuse tissue for small intestine anastomosis. However, limited knowledge exists on its application in mucosa-mucosa end-to-end anastomosis. This study investigates the effects of initial compression pressure, out-put power, and duration time on anastomosis strength ex vivo in mucosa-mucosa end-to-end anastomosis. Methods: Ex vivo porcine bowel segments were used to create 140 mucosa-mucosa end-to-end fusions. Different experimental parameters were employed for fusion, including initial com-pression pressure (50kPa-400 kPa), output power (90W, 110W, and 140W), and fusion time (5, 10, 15, 20 s). The fusion quality was measured by burst pressure and optical microscopes. Results: The best fusion quality was achieved with an initial compressive pressure between 200 and 250 kPa, an output power of 140W, and a fusion time of 15 s. However, an increase in output power and duration time resulted in a wider range of thermal damage. There was no significant difference between the burst pressure at 15 and 20 s (p > 0.05). However, a substantial increase in thermal damage was observed with longer fusion times of 15 and 20 s (p < 0.05). Conclusion: The best fusion quality for mucosa-mucosa end-to-end anastomosis ex vivo is achieved when the initial compressive pressure is between 200 and 250 kPa, the output power is approximately 140W, and the fusion time is approximately 15 s. These findings can serve as a valuable theoretical foundation and technical guidance for conducting animal experiments in vivo and subsequent tissue regeneration.

Benchmarking a New Circular Cone-based Radiosurgery System against Clinically Tested Radiosurgery System on the same Novel Digital Linear Accelerator Platform

Objective

To examine the dosimetric characteristics of circular cones, the accuracy of dose modeling and overall treatment delivery of two radiosurgery systems integrated on a linear accelerator (Linac).

Materials and Methods

The dosimetric characteristics of circular cones (4-17.5 mm) from Varian (VC) and BrainLAB (BLC) were measured for 6 MV flattening filter free beam from Edge linac using stereotactic field diode and 0.65 cc ionization chamber following established protocols. The Eclipse and iPlan modeled dose distribution for VCs and BLCs were validated with EBT3-film measurement. End-to-end tests were performed using stereotactic phantom having PTW 60008 diode connected to a Dose-1 electrometer.

Results

The depth at dose maximum, TRP2010 and dose at 10cm depth of the same size VC and BLC agree within ± 0.7 mm, ± 0.71% and ± 0.81% respectively. Full width at half maximum (FWHM) of any cone beyond 15 mm depth increases at 1% of nominal cone size per 10 mm depth. The penumbra of 4mm and 17.5mm VC at 15 mm depth was 1.1 mm and 1.50 mm. At 300 mm depth, penumbra increased by around 0.4 mm for 4 mm cone and up to 1 mm for cone size ≥12.5 mm. The VCs penumbra values were within ±1mm of the corresponding BLCs. Scatter factors for VCs varies from 0.609 to 0.841 and were within ± 1.0% of corresponding values of BLCs. Agreement between the Eclipse and iPlan computed dose fluence and the EBT3-film measured dose fluence was >98% (γ: 1%@1 mm), and the absolute dose difference was ≤ 2.2%, except for the 4 mm cone in which it was >96% and ≤4.83%. Target localization using cone-beam computed tomography was accurate within ± 0.8 mm and ± 0.3° in translation and rotation. The end-to-end dose delivery accuracy for both radiosurgery systems was within ± 3.62%.

Conclusion

The dosimetric characteristics of Varian and BLC cones of same diameter was comparable. Both Eclipse and iPlan cone planning system modeled dose fluences agree well with the EBT3 film measurement. The end-to-end tests revealed an excellent target localization accuracy of Edge linac with satisfactory and comparable absolute dose agreement between Varian and BLC radiosurgery systems and hence these can be interchanged on edge linac.

Recent Advances in the Multistep Continuous Preparation of Apis and Fine Chemicals

Over the last two decades, with the advent of continuous flow technologies, continuous processes have emerged as a major area in organic synthesis. In this context, continuous flow processes have been increasing in the preparation of Active Pharmaceutical Ingredients (APIs) and fine chemicals, such as complex synthetic intermediates, agrochemicals, and fragrances. Thus, the development of multi-step protocols has attracted special interest from the academic and industrial chemistry communities. In addition to the beneficial aspects intrinsically associated with continuous processes (e.g., waste reduction, optimal heat transfer, improved safety, and the possibility to work under harsh reaction conditions and with more dangerous reagents), these protocols also allow a rapid increase in molecular complexity. Moreover, in telescoped multi-step processes, isolation and purification steps are generally avoided or, if necessary, carried out in-line, presenting an important economy of time, solvents, reagents, and labor. Last, important synthetic strategies such as photochemical and electrochemical reactions are compatible with flow processes and are delivering relevant advances to the synthetic approaches. In this review, a general overview of the fundamentals of continuous flow processes is presented. Recent examples of multi-step continuous processes for the preparation of fine chemicals, including telescoped and end-to-end processes, are discussed, pointing out the possible advantages and/or limitations of each of these methodologies.

MR-linac daily semi-automated end-to-end quality control verification

PURPOSE

Adaptive radiation therapy (ART) on the integrated Elekta Unity magnetic resonance (MR)-linac requires routine quality assurance to verify delivery accuracy and system data transfer. In this work, our objective was to develop and validate a novel automated end-to-end test suite that verifies data transfer between multiple software platforms and quantifies the performance of multiple machine subcomponents critical to the ART process.

METHODS

We designed and implemented a software tool to quantify the MR and megavoltage (MV) isocenter coincidence, treatment couch positioning consistency, isocenter shift accuracy for the adapted plan as well as the MLC and jaw position accuracy following the beam aperture adaptation. Our tool employs a reference treatment plan with a simulated isocenter shift generated on an MR image of a readily available phantom with MR and MV visible fiducials. Execution of the test occurs within the standard adapt-to-position (ATP) clinical workflow with MV images collected of the delivered treatment fields. Using descriptive statistics, we quantified uncertainty in couch positioning, isocentre shift as well as the jaw and MLC positions of the adapted fields. We also executed sensitivity measurements to evaluate the detection algorithm's performance.

RESULTS

We report the results of 301 daily testing instances. We demonstrated consistent tracking of the MR-to-MV alignment with respect to the established value and to detect small changes on the order of 0.2 mm following machine service events. We found couch position consistency relative to the test baseline value was within 95% CI [-0.31, 0.26 mm]. For phantom shifts that form the basis for the plan adaptation, we found agreement between MV-image-detected phantom shift and online image registration, within ± 1.5 mm in all directions with a 95% CI difference of [-1.29, 0.79 mm]. For beam aperture adaptation accuracy, we found differences between the planned and detected jaw positions had a mean value of 0.27 mm and 95% CI of [-0.29, 0.82 mm] and -0.17 mm and 95% CI of [-0.37, 0.05 mm] for the MLC positions. Automated fiducial detected accuracy was within 0.08 ± 0.20 mm of manual localization. Introduced jaw and MLC position errors (1-10 mm) were detected within 0.55 mm (within 1 mm for 15/256 instances for the jaws). Phantom shifts (1.3 or 5 mm in each cardinal direction) from a reference position were detected within 0.26 mm.

CONCLUSIONS

We have demonstrated the accuracy and sensitivity of a daily end-to-end test suite capable of detecting errors in multiple machine subcomponents including system data transfer. Our test suite evaluates the entire treatment workflow and has captured system communication issues prior to patient treatment. With automated processing and the use of a standard vendor-provided phantom, it is possible to expand to other Unity sites.

ResSKNet-SSDP: Effective and Light End-To-End Architecture for Speaker Recognition

In speaker recognition tasks, convolutional neural network (CNN)-based approaches have shown significant success. Modeling the long-term contexts and efficiently aggregating the information are two challenges in speaker recognition, and they have a critical impact on system performance. Previous research has addressed these issues by introducing deeper, wider, and more complex network architectures and aggregation methods. However, it is difficult to significantly improve the performance with these approaches because they also have trouble fully utilizing global information, channel information, and time-frequency information. To address the above issues, we propose a lighter and more efficient CNN-based end-to-end speaker recognition architecture, ResSKNet-SSDP. ResSKNet-SSDP consists of a residual selective kernel network (ResSKNet) and self-attentive standard deviation pooling (SSDP). ResSKNet can capture long-term contexts, neighboring information, and global information, thus extracting a more informative frame-level. SSDP can capture short- and long-term changes in frame-level features, aggregating the variable-length frame-level features into fixed-length, more distinctive utterance-level features. Extensive comparison experiments were performed on two popular public speaker recognition datasets, Voxceleb and CN-Celeb, with current state-of-the-art speaker recognition systems and achieved the lowest EER/DCF of 2.33%/0.2298, 2.44%/0.2559, 4.10%/0.3502, and 12.28%/0.5051. Compared with the lightest x-vector, our designed ResSKNet-SSDP has 3.1 M fewer parameters and 31.6 ms less inference time, but 35.1% better performance. The results show that ResSKNet-SSDP significantly outperforms the current state-of-the-art speaker recognition architectures on all test sets and is an end-to-end architecture with fewer parameters and higher efficiency for applications in realistic situations. The ablation experiments further show that our proposed approaches also provide significant improvements over previous methods.

Artificial Intelligence-Enabled End-To-End Detection and Assessment of Alzheimer's Disease Using Voice

There is currently no simple, widely available screening method for Alzheimer's disease (AD), partly because the diagnosis of AD is complex and typically involves expensive and sometimes invasive tests not commonly available outside highly specialized clinical settings. Here, we developed an artificial intelligence (AI)-powered end-to-end system to detect AD and predict its severity directly from voice recordings. At the core of our system is the pre-trained data2vec model, the first high-performance self-supervised algorithm that works for speech, vision, and text. Our model was internally evaluated on the ADReSSo (Alzheimer's Dementia Recognition through Spontaneous Speech only) dataset containing voice recordings of subjects describing the Cookie Theft picture, and externally validated on a test dataset from DementiaBank. The AI model can detect AD with average area under the curve (AUC) of 0.846 and 0.835 on held-out and external test set, respectively. The model was well-calibrated (Hosmer-Lemeshow goodness-of-fit p-value = 0.9616). Moreover, the model can reliably predict the subject's cognitive testing score solely based on raw voice recordings. Our study demonstrates the feasibility of using the AI-powered end-to-end model for early AD diagnosis and severity prediction directly based on voice, showing its potential for screening Alzheimer's disease in a community setting.

ECG-COVID: An End-to-End Deep Model Based on Electrocardiogram for COVID-19 Detection

The early and accurate detection of COVID-19 is vital nowadays to avoid the vast and rapid spread of this virus and ease lockdown restrictions. As a result, researchers developed methods to diagnose COVID-19. However, these methods have several limitations. Therefore, presenting new methods is essential to improve the diagnosis of COVID-19. Recently, investigation of the electrocardiogram (ECG) signals becoming an easy way to detect COVID-19 since the ECG process is non-invasive and easy to use. Therefore, we proposed in this paper a novel end-to-end deep learning model (ECG-COVID) based on ECG for COVID-19 detection. We employed several deep Convolutional Neural Networks (CNNs) on a dataset of 1109 ECG images, which is built for screening the perception of COVID-19 and cardiac patients. After that, we selected the most efficient model as our model for evaluation. The proposed model is end-to-end where the input ECG images are fed directly to the model for the final decision without using any additional stages. The proposed method achieved an average accuracy of 98.81%, Precision of 98.8%, Sensitivity of 98.8% and, F1-score of 98.81% for COVID-19 detection. As cases of corona continue to rise and hospitalizations continue again, hospitals may find our study helpful when dealing with these patients who did not get significantly worse.

LRF-WiVi: A WiFi and Visual Indoor Localization Method Based on Low-Rank Fusion

In this paper, a WiFi and visual fingerprint localization model based on low-rank fusion (LRF-WiVi) is proposed, which makes full use of the complementarity of heterogeneous signals by modeling both the signal-specific actions and interaction of location information in the two signals end-to-end. Firstly, two feature extraction subnetworks are designed to extract the feature vectors containing location information of WiFi channel state information (CSI) and multi-directional visual images respectively. Then, the low-rank fusion module efficiently aggregates the specific actions and interactions of the two feature vectors while maintaining low computational complexity. The fusion features obtained are used for position estimation; In addition, for the CSI feature extraction subnetwork, we designed a novel construction method of CSI time-frequency characteristic map and a double-branch CNN structure to extract features. LRF-WiVi jointly learns the parameters of each module under the guidance of the same loss function, making the whole model more consistent with the goal of fusion localization. Extensive experiments are conducted in a complex laboratory and an open hall to verify the superior performance of LRF-WiVi in utilizing WiFi and visual signal complementarity. The results show that our method achieves more advanced positioning performance than other methods in both scenarios.

MVI-Mind: A Novel Deep-Learning Strategy Using Computed Tomography (CT)-Based Radiomics for End-to-End High Efficiency Prediction of Microvascular Invasion in Hepatocellular Carcinoma

Simple Summary

Microvascular invasion is an important indicator for reflecting the prognosis of hepatocellular carcinoma, but the traditional diagnosis requires a postoperative pathological examination. This study is the first to propose an end-to-end deep learning architecture for predicting microvascular invasion in hepatocellular carcinoma by collecting retrospective data. This method can achieve noninvasive, accurate and efficient preoperative prediction only through the patient’s radiomic data, which is very beneficial to doctors for clinical decision making in HCC patients.

Abstract

Microvascular invasion (MVI) in hepatocellular carcinoma (HCC) directly affects a patient’s prognosis. The development of preoperative noninvasive diagnostic methods is significant for guiding optimal treatment plans. In this study, we investigated 138 patients with HCC and presented a novel end-to-end deep learning strategy based on computed tomography (CT) radiomics (MVI-Mind), which integrates data preprocessing, automatic segmentation of lesions and other regions, automatic feature extraction, and MVI prediction. A lightweight transformer and a convolutional neural network (CNN) were proposed for the segmentation and prediction modules, respectively. To demonstrate the superiority of MVI-Mind, we compared the framework’s performance with that of current, mainstream segmentation, and classification models. The test results showed that MVI-Mind returned the best performance in both segmentation and prediction. The mean intersection over union (mIoU) of the segmentation module was 0.9006, and the area under the receiver operating characteristic curve (AUC) of the prediction module reached 0.9223. Additionally, it only took approximately 1 min to output a prediction for each patient, end-to-end using our computing device, which indicated that MVI-Mind could noninvasively, efficiently, and accurately predict the presence of MVI in HCC patients before surgery. This result will be helpful for doctors to make rational clinical decisions.

Automatic and Efficient Prediction of Hematoma Expansion in Patients with Hypertensive Intracerebral Hemorrhage Using Deep Learning Based on CT Images

Patients with hypertensive intracerebral hemorrhage (ICH) have a high hematoma expansion (HE) incidence. Noninvasive prediction HE helps doctors take effective measures to prevent accidents. This study retrospectively analyzed 253 cases of hypertensive intraparenchymal hematoma. Baseline non-contrast-enhanced CT scans (NECTs) were collected at admission and compared with subsequent CTs to determine the presence of HE. An end-to-end deep learning method based on CT was proposed to automatically segment the hematoma region, region of interest (ROI) feature extraction, and HE prediction. A variety of algorithms were employed for comparison. U-Net with attention performs best in the task of segmenting hematomas, with the mean Intersection overUnion (mIoU) of 0.9025. ResNet-34 achieves the most robust generalization capability in HE prediction, with an area under the receiver operating characteristic curve (AUC) of 0.9267, an accuracy of 0.8827, and an F1 score of 0.8644. The proposed method is superior to other mainstream models, which will facilitate accurate, efficient, and automated HE prediction.

Automatic Walking Method of Construction Machinery Based on Binocular Camera Environment Perception

In this paper, we propose an end-to-end automatic walking system for construction machinery, which uses binocular cameras to capture images of construction machinery for environmental perception, detects target information in binocular images, estimates the relative distance between the current target and cameras, and predicts the real-time control signal of construction machinery. This system consists of two parts: the binocular recognition ranging model and the control model. Objects within 5 m can be quickly detected by the recognition ranging model, and at the same time, the distance of the object can be accurately ranged to ensure the full perception of the surrounding environment of the construction machinery. The distance information of the object, the feature information of the binocular image, and the control signal of the previous stage are sent to the control model; then, the prediction of the control signal of the construction machinery can be output in the next stage. In this way, the automatic walking experiment of the construction machinery in a specific scenario is completed, which proves that the model can control the machinery to complete the walking task smoothly and safely.

Ileocolic Resection for Crohn Disease: The Influence of Different Surgical Techniques on Perioperative Outcomes, Recurrence Rates, and Endoscopic Surveillance

Ileocolic resection (ICR) is the most common surgical procedure in Crohn disease (CD). There are many surgical techniques for performing ICRs and subsequent anastomoses. Recurrence of CD after ICR is common, often clinically silent, and thus requires monitoring including periodic use of endoscopy to detect early active disease. There is emerging evidence that surgical approaches may influence CD recurrence. This review explores the various surgical considerations, the data behind each decision, and how these techniques influence subsequent endoscopic surveillance.

An End-to-End Depression Recognition Method Based on EEGNet

Major depressive disorder (MDD) is a common and highly debilitating condition that threatens the health of millions of people. However, current diagnosis of depression relies on questionnaires that are highly correlated with physician experience and hence not completely objective. Electroencephalography (EEG) signals combined with deep learning techniques may be an objective approach to effective diagnosis of MDD. This study proposes an end-to-end deep learning framework for MDD diagnosis based on EEG signals. We used EEG signals from 29 healthy subjects and 24 patients with severe depression to calculate Accuracy, Precision, Recall, F1-Score, and Kappa coefficient, which were 90.98%, 91.27%, 90.59%, and 81.68%, respectively. In addition, we found that these values were highest when happy-neutral face pairs were used as stimuli for detecting depression. Compared with exiting methods for EEG-based MDD classification, ours can maintain stable model performance without re-calibration. The present results suggest that the method is highly accurate for diagnosis of MDD and can be used to develop an automatic plug-and-play EEG-based system for diagnosing depression.

DIRECT-Net: A unified mutual-domain material decomposition network for quantitative dual-energy CT imaging

PURPOSE

The purpose of this paper is to present an end-to-end deep convolutional neural network to improve the dual-energy CT (DECT) material decomposition performance.

METHODS

In this study, we proposes a unified mutual-domain (sinogram domain and CT domain) material decomposition network (DIRECT-Net) for DECT imaging. By design, the DIRECT-Net has immediate access to mutual-domain data, and utilizes stacked convolution neural network layers for noise reduction and material decomposition. The training data are numerically generated following the fundamental DECT imaging physics. Numerical simulation of the XCAT digital phantom, experiments of a biological specimen, a calcium chloride phantom and an iodine solution phantom are carried out to evaluate the performance of DIRECT-Net. Comparisons are performed with different DECT decomposition algorithms.

RESULTS

Results demonstrate that the proposed DIRECT-Net can generate water and bone basis images with less artifacts compared to the other decomposition methods. Additionally, the quantification errors of the calcium chloride (75-375 mg/cm³ ) and the iodine (2-20 mg/cm³ ) are less than 4%.

CONCLUSIONS

An end-to-end material decomposition network is proposed for quantitative DECT imaging. The qualitative and quantitative results demonstrate that this new DIRECT-Net has promising benefits in improving the DECT image quality.

Outcomes of Tibial Nerve Repair and Transfer: A Structured Evidence-Based Systematic Review and Meta-Analysis

Although nerve transfer and repair are well-established for treatment of nerve injury in the upper extremity, there are no established parameters for when or which treatment modalities to utilize for tibial nerve injuries. The objective of our study is to conduct a systematic review of the effectiveness of end-to-end repair, neurolysis, nerve grafting, and nerve transfer in improving motor function after tibial nerve injury. PubMed, Cochrane, Medline, and Embase libraries were queried according to the PRISMA guidelines for articles that present functional outcomes after tibial nerve injury in humans treated with nerve transfer or repair. The final selection included Nineteen studies with 677 patients treated with neurolysis (373), grafting (178), end-to-end repair (90), and nerve transfer (30), from 1985 to 2018. The mean age of all patients was 27.0 ± 10.8 years, with a mean preoperative interval of 7.4 ± 10.5 months, and follow-up period of 82.9 ± 25.4 months. The mean graft repair length for nerve transfer and grafting patients was 10.0 ± 5.8 cm, and the most common donor nerve was the sural nerve. The most common mechanism of injury was gunshot wound, and the mean MRC of all patients was 3.7 ± 0.6. Good outcomes were defined as MRC ≥ 3. End-to-end repair treatment had the greatest number of good outcomes, followed by neurolysis. Patients with preoperative intervals less than 7 months were more likely to have good outcomes than those greater than 7 months. Patients with sport injuries had the highest percentage of good outcomes in contrast to patients with transections and who were in MVAs. We found no statistically significant difference in good outcomes between the use of sural and peroneal donor nerve grafts, nor between age, graft length, and MRC score.

Long-Term Ship Position Prediction Using Automatic Identification System (AIS) Data and End-to-End Deep Learning

The establishment of maritime safety and security is an important concern. Ship position prediction for maritime situational awareness (MSA), as a critical aspect of maritime safety and security, requires a longer time interval than collision avoidance and maritime traffic monitoring. However, previous studies focused mainly on shorter time-interval predictions ranging from 30 min to 10 h. A longer time-interval ship position prediction is required not only for MSA, but also for efficient allocation of ships by shipping companies in accordance with global freight demand. This study used an end-to-end tracking method that inputs the previous position of a vessel to a trained deep learning model to predict its next position with an average 24-h interval. An AIS dataset with a long-time-interval distribution in a nine-year timespan for capesize bulk carriers worldwide was used. In the first experiment, a deep learning model of the Indian Ocean was examined. Subsequently, the model performance was compared for six different oceans and six primary maritime chokepoints to investigate the influence of each area. In the third experiment, a sample location within the Malacca Strait area was selected, and the number of ships was counted daily. The results indicate that the ship position can be predicted accurately with an average time interval of 24 h using deep learning systems with AIS data.

Mechanical Stimulation after Centrifuge-Free Nano-Electroporative Transfection Is Efficient and Maintains Long-Term T Cell Functionalities

Transfection is an essential step in genetic engineering and cell therapies. While a number of non-viral micro- and nano-technologies have been developed to deliver DNA plasmids into the cell cytoplasm, one of the most challenging and least efficient steps is DNA transport to and expression in the nucleus. Here, the magnetic nano-electro-injection (MagNEI) platform is described which makes use of oscillatory mechanical stimulation after cytoplasmic delivery with high aspect-ratio nano-structures to achieve stable (>2 weeks) net transfection efficiency (efficiency × viability) of 50% in primary human T cells. This is, to the best of the authors' knowledge, the highest net efficiency reported for primary T cells using a centrifuge-free, non-viral transfection method, in the absence of cell selection, and with a clinically relevant cargo size (>12 kbp). Wireless mechanical stimulation downregulates the expression of microtubule motor protein gene, KIF2A, which increases local DNA concentration near the nuclei, resulting in enhanced DNA transfection. Magnetic forces also accelerate membrane repair by promoting actin cytoskeletal remodeling which preserves key biological attributes including cell proliferation and gene expressions. These results demonstrate MagNEI as a powerful non-viral transfection technique for progress toward fully closed, end-to-end T cell manufacturing with less human labor, lower production cost, and shorter delay.

Scale-Invariant Multidirectional License Plate Detection with the Network Combining Indirect and Direct Branches

As the license plate is multiscale and multidirectional in the natural scene image, its detection is challenging in many applications. In this work, a novel network that combines indirect and direct branches is proposed for license plate detection in the wild. The indirect detection branch performs small-sized vehicle plate detection with high precision in a coarse-to-fine scheme using vehicle–plate relationships. The direct detection branch detects the license plate directly in the input image, reducing false negatives in the indirect detection branch due to the miss of vehicles’ detection. We propose a universal multidirectional license plate refinement method by localizing the four corners of the license plate. Finally, we construct an end-to-end trainable network for license plate detection by combining these two branches via post-processing operations. The network can effectively detect the small-sized license plate and localize the multidirectional license plate in real applications. To our knowledge, the proposed method is the first one that combines indirect and direct methods into an end-to-end network for license plate detection. Extensive experiments verify that our method outperforms the indirect methods and direct methods significantly.

Pilot-scale demonstration of an end-to-end integrated and continuous biomanufacturing process

There has been increasing momentum recently in the biopharmaceutical industry to transition from traditional batch processes to next-generation integrated and continuous biomanufacturing. This transition from batch to continuous is expected to offer several advantages which, taken together, could significantly improve access to biologics drugs for patients. Despite this recent momentum, there has not been a commercial implementation of a continuous bioprocess reported in the literature. In this study, we describe a successful pilot-scale proof-of-concept demonstration of an end-to-end integrated and continuous bioprocess for the production of a monoclonal antibody (mAb). This process incorporated all of the key unit operations found in a typical mAb production process, including the final steps of virus removal filtration, ultrafiltration, diafiltration, and formulation. The end-to-end integrated process was operated for a total of 25 days and produced a total of 4.9 kg (200 g/day or 2 g/L BRX/day) of the drug substance from a 100-L perfusion bioreactor (BRX) with acceptable product quality and minimal operator intervention. This successful proof-of-concept demonstrates that end-to-end integrated continuous bioprocessing is achievable with current technologies and represents an important step toward the realization of a commercial integrated and continuous bioprocessing process.

Coarse-to-Fine Satellite Images Change Detection Framework via Boundary-Aware Attentive Network

Timely and accurate change detection on satellite images by using computer vision techniques has been attracting lots of research efforts in recent years. Existing approaches based on deep learning frameworks have achieved good performance for the task of change detection on satellite images. However, under the scenario of disjoint changed areas in various shapes on land surface, existing methods still have shortcomings in detecting all changed areas correctly and representing the changed areas boundary. To deal with these problems, we design a coarse-to-fine detection framework via a boundary-aware attentive network with a hybrid loss to detect the change in high resolution satellite images. Specifically, we first perform an attention guided encoder-decoder subnet to obtain the coarse change map of the bi-temporal image pairs, and then apply residual learning to obtain the refined change map. We also propose a hybrid loss to provide the supervision from pixel, patch, and map levels. Comprehensive experiments are conducted on two benchmark datasets: LEBEDEV and SZTAKI to verify the effectiveness of the proposed method and the experimental results show that our model achieves state-of-the-art performance.

End-to-End Protocol for the Detection of SARS-CoV-2 from Built Environments

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes coronavirus disease 2019, is a respiratory virus primarily transmitted person to person through inhalation of droplets or aerosols, laden with viral particles. However, as recent studies have shown, virions can remain infectious for up to 72 h on surfaces, which can lead to transmission through contact. Thus, a comprehensive study was conducted to determine the efficiency of protocols to recover SARS-CoV-2 from surfaces in built environments. This end-to-end (E2E) study showed that the effective combination for monitoring SARS-CoV-2 on surfaces includes using an Isohelix swab collection tool, DNA/RNA Shield as a preservative, an automated system for RNA extraction, and reverse transcriptase quantitative PCR (RT-qPCR) as the detection assay. Using this E2E approach, this study showed that, in some cases, noninfectious viral fragments of SARS-CoV-2 persisted on surfaces for as long as 8 days even after bleach treatment. Additionally, debris associated with specific built environment surfaces appeared to inhibit and negatively impact the recovery of RNA; Amerstat demonstrated the highest inhibition (>90%) when challenged with an inactivated viral control. Overall, it was determined that this E2E protocol required a minimum of 1,000 viral particles per 25 cm2 to successfully detect virus from test surfaces. Despite our findings of viral fragment longevity on surfaces, when this method was employed to evaluate 368 samples collected from various built environmental surfaces, all samples tested negative, indicating that the surfaces were either void of virus or below the detection limit of the assay.IMPORTANCE The ongoing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (the virus responsible for coronavirus disease 2019 [COVID-19]) pandemic has led to a global slowdown with far-reaching financial and social impacts. The SARS-CoV-2 respiratory virus is primarily transmitted from person to person through inhalation of infected droplets or aerosols. However, some studies have shown that virions can remain infectious on surfaces for days and can lead to human infection from contact with infected surfaces. Thus, a comprehensive study was conducted to determine the efficiency of protocols to recover SARS-CoV-2 from surfaces in built environments. This end-to-end study showed that the effective combination for monitoring SARS-CoV-2 on surfaces required a minimum of 1,000 viral particles per 25 cm2 to successfully detect virus from surfaces. This comprehensive study can provide valuable information regarding surface monitoring of various materials as well as the capacity to retain viral RNA and allow for effective disinfection.

A single blinded randomized controlled trial comparing semi-mechanical with hand-sewn cervical anastomosis after esophagectomy for cancer (SHARE-study)

OBJECTIVE

The aim was to compare leak rate between hand-sewn end-to-end anastomosis (ETE) and semi-mechanical anastomosis (SMA) after esophagectomy with gastric tube reconstruction.

BACKGROUND DATA

The optimal surgical technique for creation of an anastomosis in the neck after esophagectomy is unclear.

METHODS

Patients with esophageal cancer undergoing esophagectomy with gastric tube reconstruction and cervical anastomosis were eligible for participation after written informed consent. Patients were randomized in 1:1 ratio. Primary endpoint was anastomotic leak rate defined as external drainage of saliva from the site of the anastomosis or intra-thoracic manifestation of leak. Secondary endpoints included anastomotic stricture rate at one year follow up, number of endoscopic dilatations, dysphagia-score, hospital stay, morbidity, and mortality. Patients were blinded for intervention.

RESULTS

Between August 2011 and July 2014, 174 patients with esophageal cancer underwent esophagectomy. Ninety-three patients were randomized to ETE (n = 44) or SMA (n = 49). Anastomotic leak occurred in 9 of 44 patients (20%) in the ETE group and 12 of 49 patients (24%) in the SMA group (absolute difference 4%, 95% CI -13% to +21%; p = .804). There was no significant difference in dysphagia at 1 year postoperatively (ETE 25% vs. SMA 20%; p = .628), in stricture rate (ETE 25% vs. 19% in SMA, p = .46), nor in median hospital stay (17 days in the ETE group, 13 days in the SMA group), morbidity (82% vs. 73%, p = .460) or mortality (0% vs. 4%, p = .175) between the groups.

Ophthalmologist-Level Classification of Fundus Disease With Deep Neural Networks

Purpose

To implement the classification of fundus diseases using deep convolutional neural networks (CNN), which is trained end-to-end from fundus images directly, the only input are pixels and disease labels, and the output is a probability distribution of a fundus image belonging to 18 fundus diseases.

Methods

Automated classification of fundus diseases using images is a challenging task owing to the fine-grained variability in the appearance of fundus lesions. Deep CNNs show potential for general and highly variable tasks across many fine-grained object categories. Deep CNNs need large amounts of labeled samples, yet the available fundus images, especially labeled samples, are limited, which cannot satisfy the training requirement. So image augmentations such as rotation, scaling, and noising are implemented to enlarge the training dataset. We fine-tune the ResNet CNN architecture with 120,100 fundus images consisting of 18 different diseases and use it to classify the fundus images into corresponding diseases.

Results

The performance is tested against two board-certified ophthalmologists. The CNN achieves performance on par with the experts for the classification accuracy.

Conclusions

Deep CNN is capable of predicting fundus diseases given fundus images as input, which can enhance the efficiency of diagnosis process and promote better visual outcomes. Outfitted with deep neural networks, mobile devices can potentially extend the reach of ophthalmologists outside of the clinic and provide low-cost universal access to vital diagnostic care.

Translational Relevance

This article implemented automatic prediction of fundus diseases that was done by ophthalmologists previously.

End-to-End Training for Compound Expression Recognition

For a long time, expressions have been something that human beings are proud of. That is an essential difference between us and machines. With the development of computers, we are more eager to develop communication between humans and machines, especially communication with emotions. The emotional growth of computers is similar to the growth process of each of us, starting with a natural, intimate, and vivid interaction by observing and discerning emotions. Since the basic emotions, angry, disgusted, fearful, happy, neutral, sad and surprised are put forward, there are many researches based on basic emotions at present, but few on compound emotions. However, in real life, people’s emotions are complex. Single expressions cannot fully and accurately show people’s inner emotional changes, thus, exploration of compound expression recognition is very essential to daily life. In this paper, we recommend a scheme of combining spatial and frequency domain transform to implement end-to-end joint training based on model ensembling between models for appearance and geometric representations learning for the recognition of compound expressions in the wild. We are mainly devoted to digging the appearance and geometric information based on deep learning models. For appearance feature acquisition, we adopt the idea of transfer learning, introducing the ResNet50 model pretrained on VGGFace2 for face recognition to implement the fine-tuning process. Here, we try and compare two minds, one is that we utilize two static expression databases FER2013 and RAF Basic for basic emotion recognition to fine tune, the other is that we fine tune the model on the input three channels composed of images generated by DWT2 and WAVEDEC2 wavelet transforms based on rbio3.1 and sym1 wavelet bases respectively. For geometric feature acquisition, we firstly introduce a densesift operator to extract facial key points and their histogram descriptions. After that, we introduce deep SAE with a softmax function, stacked LSTM and Sequence-to-Sequence with stacked LSTM and define their structures by ourselves. Then, we feed the salient key points and their descriptions into three models to train respectively and compare their performances. When the model training for appearance and geometric features learning is completed, we combine the two models with category labels to achieve further end-to-end joint training, considering that ensembling models, which describe different information, can further improve recognition results. Finally, we validate the performance of our proposed framework on an RAF Compound database and achieve a recognition rate of 66.97%. Experiments show that integrating different models, which express different information, and achieving end-to-end training can quickly and effectively improve the performance of the recognition.

End-to-End Automatic Pronunciation Error Detection Based on Improved Hybrid CTC/Attention Architecture

Advanced automatic pronunciation error detection (APED) algorithms are usually based on state-of-the-art automatic speech recognition (ASR) techniques. With the development of deep learning technology, end-to-end ASR technology has gradually matured and achieved positive practical results, which provides us with a new opportunity to update the APED algorithm. We first constructed an end-to-end ASR system based on the hybrid connectionist temporal classification and attention (CTC/attention) architecture. An adaptive parameter was used to enhance the complementarity of the connectionist temporal classification (CTC) model and the attention-based seq2seq model, further improving the performance of the ASR system. After this, the improved ASR system was used in the APED task of Mandarin, and good results were obtained. This new APED method makes force alignment and segmentation unnecessary, and it does not require multiple complex models, such as an acoustic model or a language model. It is convenient and straightforward, and will be a suitable general solution for L1-independent computer-assisted pronunciation training (CAPT). Furthermore, we find that find that in regards to accuracy metrics, our proposed system based on the improved hybrid CTC/attention architecture is close to the state-of-the-art ASR system based on the deep neural network-deep neural network (DNN-DNN) architecture, and has a stronger effect on the F-measure metrics, which are especially suitable for the requirements of the APED task.

An Autonomous Path Planning Model for Unmanned Ships Based on Deep Reinforcement Learning

Deep reinforcement learning (DRL) has excellent performance in continuous control problems and it is widely used in path planning and other fields. An autonomous path planning model based on DRL is proposed to realize the intelligent path planning of unmanned ships in the unknown environment. The model utilizes the deep deterministic policy gradient (DDPG) algorithm, through the continuous interaction with the environment and the use of historical experience data; the agent learns the optimal action strategy in a simulation environment. The navigation rules and the ship's encounter situation are transformed into a navigation restricted area, so as to achieve the purpose of planned path safety in order to ensure the validity and accuracy of the model. Ship data provided by ship automatic identification system (AIS) are used to train this path planning model. Subsequently, the improved DRL is obtained by combining DDPG with the artificial potential field. Finally, the path planning model is integrated into the electronic chart platform for experiments. Through the establishment of comparative experiments, the results show that the improved model can achieve autonomous path planning, and it has good convergence speed and stability.

End-to-end test and MOSFET in vivo skin dosimetry for 192Ir high-dose-rate brachytherapy of chronic psoriasis

Purpose

This study was performed using end-to-end testing and real-time in vivo skin dose measurements, using metal oxide semiconductor field effect transistor (MOSFET) dosimeters on our first chronic psoriasis patient treated with iridium-192 (¹⁹²Ir) high-dose-rate (HDR) brachytherapy (BT).

Material and methods

Treatment delivery was planned with the prescription dose of 1.8 Gy to a 3 mm depth for 12 fractions, using our custom-fabricated surface mold and Varian soft catheters. The optimal technique to provide an adequate and acceptable skin dose as well as its feasibility were evaluated by an end-to-end exercise using a perspex finger phantom. The accuracy and reliability of MOSFET dose measurement was explored with a thermoluminescence dosimetry (TLD) before being used in vivo to monitor skin doses during treatment delivery for each BT fraction.

Results

Using custom-made surface mold (2.4 mm Med-Tec thermoplastic mask for hand fixation and 5 applicators attached to each finger for dose delivery), the optimal skin dose on the phantom was obtained without the need for additional bolus to increase thickness of applicator. We acquired mean skin doses at different skin depths from various dose-volume parameters of no-bolus and 3 mm-added bolus plans. They were 125% and 110% (1 mm), 120% and 108% (2 mm), and 114% and 106% (3 mm), respectively. There was excellent agreement between MOSFET and TLD for ¹⁹²Ir HDR-BT within ±3% (mean 2.65%, SD = 2.05%). With no energy correction, MOSFET overestimated the Acuros BV surface doses by up to 7% in the phantom study and in the clinical case.

Conclusions

We demonstrated achievable HDR-BT for our first case of nail bed psoriasis. The end-to-end exercise was an efficient methodology to evaluate new feasibility for this technique. Real-time dose monitoring using MOSFET was an effective and reliable tool to ensure treatment quality and patient safety.

Management for Ureteral Stenosis: A Comparison of Robot-Assisted Laparoscopic Ureteroureterostomy and Conventional Laparoscopic Ureteroureterostomy

Background: To describe and analyze our experience of robotic-assisted laparoscopic ureteroureterostomy (RALU) and conventional LU for the repair of ureteral stenosis and compare the differences of safety and efficacy between RALU and LU. Materials and Methods: Patients who underwent RALU or LU for ureteral stenosis were retrospectively analyzed. Baseline characteristics, details of stenosis, surgery and some laboratory tests, and follow-up data were collected and analyzed. Results: Among 126 patients presented with ureteral stenosis, 65 patients underwent RALU and 61 patients underwent LU. All operations were completed successfully without conversion to open surgery. Both groups were comparable in baseline characteristics and details of stenosis. The mean operative time, suturing time, and hospitalization time of patients in RALU group were significantly less than those in LU group. The mean operative time of the RALU group was 126.34 minutes, whereas the mean operative time of the LU group was 176.57 minutes (P < .001). The average suturing time of RALU and LU was 26.88 and 70.43 minutes, respectively (P < .001). The mean hospitalization time of RALU and LU was 4.01 and 5.02, respectively (P < .001). RALU presented a lower degree of leukocytes rise than LU (P < .001). The mean follow-up time was 29.52 months. Conclusions: RALU and LU both are safe and feasible for ureteral stenosis with a low incidence of complications. Compared with LU, RALU may be a better choice with shorter operative time, suturing time, postoperative hospitalization time, and slighter inflammation. Further clinical studies of high quality are needed to confirm the priority of RALU.

Deep Autoencoder Neural Networks for Short-Term Traffic Congestion Prediction of Transportation Networks

Traffic congestion prediction is critical for implementing intelligent transportation systems for improving the efficiency and capacity of transportation networks. However, despite its importance, traffic congestion prediction is severely less investigated compared to traffic flow prediction, which is partially due to the severe lack of large-scale high-quality traffic congestion data and advanced algorithms. This paper proposes an accessible and general workflow to acquire large-scale traffic congestion data and to create traffic congestion datasets based on image analysis. With this workflow we create a dataset named Seattle Area Traffic Congestion Status (SATCS) based on traffic congestion map snapshots from a publicly available online traffic service provider Washington State Department of Transportation. We then propose a deep autoencoder-based neural network model with symmetrical layers for the encoder and the decoder to learn temporal correlations of a transportation network and predicting traffic congestion. Our experimental results on the SATCS dataset show that the proposed DCPN model can efficiently and effectively learn temporal relationships of congestion levels of the transportation network for traffic congestion forecasting. Our method outperforms two other state-of-the-art neural network models in prediction performance, generalization capability, and computation efficiency.

SeqSleepNet: End-to-End Hierarchical Recurrent Neural Network for Sequence-to-Sequence Automatic Sleep Staging

Automatic sleep staging has been often treated as a simple classification problem that aims at determining the label of individual target polysomnography epochs one at a time. In this paper, we tackle the task as a sequence-to-sequence classification problem that receives a sequence of multiple epochs as input and classifies all of their labels at once. For this purpose, we propose a hierarchical recurrent neural network named SeqSleepNet (source code is available at http://github.com/pquochuy/SeqSleepNet). At the epoch processing level, the network consists of a filterbank layer tailored to learn frequency-domain filters for preprocessing and an attention-based recurrent layer designed for short-term sequential modeling. At the sequence processing level, a recurrent layer placed on top of the learned epoch-wise features for long-term modeling of sequential epochs. The classification is then carried out on the output vectors at every time step of the top recurrent layer to produce the sequence of output labels. Despite being hierarchical, we present a strategy to train the network in an end-to-end fashion. We show that the proposed network outperforms the state-of-the-art approaches, achieving an overall accuracy, macro F1-score, and Cohen's kappa of 87.1%, 83.3%, and 0.815 on a publicly available dataset with 200 subjects.

Intelligent Land-Vehicle Model Transfer Trajectory Planning Method Based on Deep Reinforcement Learning

To address the problem of model error and tracking dependence in the process of intelligent vehicle motion planning, an intelligent vehicle model transfer trajectory planning method based on deep reinforcement learning is proposed, which is able to obtain an effective control action sequence directly. Firstly, an abstract model of the real environment is extracted. On this basis, a deep deterministic policy gradient (DDPG) and a vehicle dynamic model are adopted to jointly train a reinforcement learning model, and to decide the optimal intelligent driving maneuver. Secondly, the actual scene is transferred to an equivalent virtual abstract scene using a transfer model. Furthermore, the control action and trajectory sequences are calculated according to the trained deep reinforcement learning model. Thirdly, the optimal trajectory sequence is selected according to an evaluation function in the real environment. Finally, the results demonstrate that the proposed method can deal with the problem of intelligent vehicle trajectory planning for continuous input and continuous output. The model transfer method improves the model's generalization performance. Compared with traditional trajectory planning, the proposed method outputs continuous rotation-angle control sequences. Moreover, the lateral control errors are also reduced.

Self-organization of nanorods into ultra-long range two-dimensional monolayer end-to-end network

Highly uniform large-scale assembly of nanoscale building blocks can enable unique collective properties for practical electronic and photonic devices. We present a two-dimensional (2-D), millimeter-scale network of colloidal CdSe nanorods (NRs) in monolayer thickness through end-to-end linking. The colloidal CdSe NRs are sterically stabilized with tetradecylphosphonic acid (TDPA), and their tips are partially etched in the presence of gold chloride (AuCl3) and didecyldimethylammonium bromide (DDAB), which make them unwetted in toluene. This change in surface wetting property leads to spontaneous adsorption at the 2-D air/toluene interface. Anisotropy in both the geometry and the surface property of the CdSe NRs causes deformation of the NR/toluene/air interface, which derives capillary attraction between tips of neighboring NRs inward. As a result, the NRs confined at the interface spontaneously form a 2-D network composed of end-to-end linkages. We employ a vertical-deposition approach to maintain a consistent rate of NR supply to the interface during the assembly. The rate control turns out to be pivotal in the preparation of a highly uniform large scale 2-D network without aggregation. In addition, unprecedented control of the NR density in the network was possible by adjusting either the lift-up speed of the immersed substrate or the relative concentration of AuCl3 to DDAB. Our findings provide important design criteria for 2-D assembly of anisotropic nanobuilding blocks.