Feng-Rong Sun

In persistent patrol and online task discovery in environments with obstacles, unmanned aerial vehicle (UAV) swarms are constrained by limited battery capacity and frequent recharging disrupts patrol continuity. In comparison, unmanned ground vehicle (UGV) fleets have higher endurance and payload capacity and can serve as mobile charging platforms while executing ground-service tasks. In such collaborative scenarios, UAVs patrol along a coverage path and discover tasks online, whereas UGVs execute discovered ground tasks and provide mobile charging support. To cope with rendezvous uncertainty due to obstacle-induced detours and inefficient usage of UGV time during charging, this study proposes an energy-constrained UAV-UGV coordination framework based on adaptive anticipatory rendezvous and time-window scheduling. In particular, the adaptive anticipatory rendezvous module handles anticipatory rendezvous planning, while the time-window scheduling module models the post-rendezvous charging stage as a schedulable time window for opportunistic ground-task insertion. Simulations demonstrate that the proposed framework consistently reduces system energy consumption, completion time, and the number of emergency landings compared with three representative baselines. Moreover, a UAV-UGV prototype with AprilTag-based visual landing and post-landing mechanical correction is developed to validate the engineering feasibility of the key closed-loop process.

Speckle tracking echocardiography (STE) is effective in assessing cardiac dysfunction in high-risk populations for heart failure. As a fundamental STE technology, speckle tracking methods have been studied for nearly two decades. Over the past decade, the rapid development of AI has led to increasing interest in deep learning-based speckle tracking methods which require large annotated datasets for training. However, the challenge of manually annotating speckle regions in echocardiograms hinders the clinical deployment of those deep learning methods, thereby maintaining the indispensability of traditional speckle tracking methods. This paper presents a clinically applicable speckle tracking method, bidirectional block matching (BiDiBM), which introduces several novel processes to enhance the tracking accuracy and robustness of traditional BM methods. We evaluated BiDiBM for myocardial longitudinal strain (MLS) on an open-source synthetic echocardiographic dataset across four scenarios, comparing its strain measurements against the gold standard using the root mean square error (RMSE) and the zero-lag point (ZERO-LAG) of cross-correlation functions. Mean RMSE values across scenarios ranged from 1.0576 ± 0.2734% to 1.2812 ± 0.4703%, with corresponding ZERO-LAG values ranging from 0.8889 ± 0.1534 to 0.9484 ± 0.0399. In comparison experiments, BiDiBM achieved superior accuracy and efficiency over the conventional BM and the deep-learning comparators. Additionally, we performed a small real-world validation, in which BiDiBM tracked speckle points, and the derived MLS curves had physiologically consistent morphologies and trends on expert review. Overall, results indicate the BiDiBM method exhibits robust accuracy and reliability, thereby confirming its suitability for clinical STE applications and establishing a practical foundation for deploying deep learning-based methods in clinical settings.

Visual impairment or even loss seriously affects quality of life. Benefited by the rapid development of sound/laser detection, Global Positioning System (GPS)/Beidou positioning, machine vision and other technologies, the quality of life of blind people is expected to be improved through visual substitution technology. The existing visual substitution devices still have limitations in terms of safety, robustness, and ease of operation. The remote companion system developed here fully utilizes multimodal navigation and remote communication technologies, and the positioning and interaction functions of commercial mobile phones. Together with the accumulated judgment of backend personnel, it can provide real-time, safe, and reliable navigation services for blind people, helping them complete daily activities such as independent travel, circulation, and shopping. The practical results show that the system not only has strong operability and is easy to use, but also can provide users with a strong sense of security and companionship, making it suitable for promotion. In the future, this system can also be promoted for other vulnerable groups such as the elderly.

Echocardiogram analysis plays a crucial role in assessing and diagnosing cardiac function, providing essential data to support medical diagnoses of heart disease. A key task, accurately identifying and segmenting the left ventricle (LV) in echocardiograms, remains challenging and labor-intensive. Current automated cardiac segmentation methods often lack the necessary accuracy and reproducibility, while semi-automated or manual annotations are excessively time-consuming. To address these limitations, we propose a novel segmentation framework, semi-and self-supervised learning with dual attention (SSL-DA) for echocardiogram segmentation. We start with a temporal masking network for pre-training. This network captures valuable information, such as echocardiogram periodicity. It also provides optimized initialization parameters for LV segmentation. We then employ a semi-supervised network to automatically segment the left ventricle, enhancing the model’s learning with channel and spatial attention mechanisms to capture global channel dependencies and spatial dependencies across annotations. We evaluated SSL-DA on the publicly available EchoNet-Dynamic dataset, achieving a Dice similarity coefficient of 93.34% (95% CI, 93.23–93.46%), outperforming most prior CNN-based models. To further assess the generalization ability of SSL-DA, we conducted ablation experiments on the CAMUS dataset. Experimental results confirm that SSL-DA can quickly and accurately segment the left ventricle in echocardiograms, showing its potential for robust clinical application.

X-ray ionizing radiation from Computed Tomography (CT) scanning increases cancer risk for patients, thus making sparse view CT, which diminishes X-ray exposure by lowering the number of projections, highly significant in diagnostic imaging. However, reducing the number of projections inherently degrades image quality, negatively impacting clinical diagnosis. Consequently, attaining reconstructed images that meet diagnostic imaging criteria for sparse view CT is challenging. This paper presents a novel network (CSUF), specifically designed for ultra-sparse view lung CT image reconstruction. The CSUF network consists of three cohesive components including (1) a compressed sensing-based CT image reconstruction module (VdCS module), (2) a U-shaped end-to-end network, CT-RDNet, enhanced with a self-attention mechanism, acting as the generator in a Generative Adversarial Network (GAN) for CT image restoration and denoising, and (3) a feedback loop. The VdCS module enriches CT-RDNet with enhanced features, while CT-RDNet supplies the VdCS module with prior images infused with rich details and minimized artifacts, facilitated by the feedback loop. Engineering simulation experimental results demonstrate the robustness of the CSUF network and its potential to deliver lung CT images with diagnostic imaging quality even under ultra-sparse view conditions.

Computer-aided diagnosis (CAD) has already been widely used in medical image processing. We recently make another trial to implement convolutional neural network (CNN) on the classification of pulmonary nodules of thoracic CT images. The biggest challenge in medical image classification with the help of CNN is the difficulty of acquiring enough samples, and overfitting is a common problem when there are not enough images for training. Transfer learning has been verified as reasonable in dealing with such problems with an acceptable loss value. We use the classic LeNet-5 model to classify pulmonary nodules of thoracic CT images, including benign and malignant pulmonary nodules, and different malignancies of the malignant nodules. The CT images are obtained from Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) where both pulmonary nodule scanning and nodule annotations are available. These images are labeled and stored in a medical images knowledge base (KB), which is designed and implemented in our previous work. We implement the 10-folder cross validation (CV) to testify the robustness of the classification model we trained. The result demonstrates that the transfer learning of the LeNet-5 is good for classifying pulmonary nodules of thoracic CT images, and the average values of Top-1 accuracy are 97.041% and 96.685% respectively. We believe that our work is beneficial and has potential for practical diagnosis of lung nodules.

The real-time myocardial contrast echocardiography (RT MCE) is a new echocardiography technology, which allows clinicians to noninvasively evaluate the perfusion of myocardial capillary of patients, using the quantitative analysis of RT MCE. But the accurate analysis requires tracking the position of region of interest (ROI) within the myocardial area, so as to compensate for the translation or rotation offsets, which are due to such uncontrollable factors as heart motion. We used diamond search method and Brox’s coarse-to-fine warping optical flow technique for this ROI tracking problem. We validated our methods by comparing the quantitative analysis results of RT MCE using our methods with those using Lucas & Kanade’s optical flow technique, which had been report to be accurate enough for this ROI tracking. We finally present some examples of animal experiment to show the effectiveness and the clinical application value of our ROI tracking methods.