Deep convolutional neural network (DCNN) models have been widely explored for skin disease diagnosis and some of them have achieved the diagnostic outcomes comparable or even superior to those of dermatologists. However, broad implementation of DCNN in skin disease detection is hindered by small size and data imbalance of the publically accessible skin lesion datasets. This paper proposes a novel single-model based strategy for classification of skin lesions on small and imbalanced datasets. First, various DCNNs are trained on different small and imbalanced datasets to verify that the models with moderate complexity outperform the larger models. Second, regularization DropOut and DropBlock are added to reduce overfitting and a Modified RandAugment augmentation strategy is proposed to deal with the defects of sample underrepresentation in the small dataset. Finally, a novel Multi-Weighted New Loss (MWNL) function and an end-to-end cumulative learning strategy (CLS) are introduced to overcome the challenge of uneven sample size and classification difficulty and to reduce the impact of abnormal samples on training. By combining Modified RandAugment, MWNL and CLS, our single DCNN model method achieved the classification accuracy comparable or superior to those of multiple ensembling models on different dermoscopic image datasets. Our study shows that this method is able to achieve a high classification performance at a low cost of computational resources and inference time, potentially suitable to implement in mobile devices for automated screening of skin lesions and many other malignancies in low resource settings.

In this brief, we focus on the offline route planning of unmanned aerial vehicle (UAV) for the coverage search mission in a river region. Given the prior likelihood distribution of area importance, this brief aims to generate UAV feasible routes which maximize the cumulative probability of finding a single and stationary target within the required time. First, Gaussian mixture model is used to approximate the prior likelihood distribution, and several river segments with high detection probability corresponding to Gaussian components can be extracted. With the consideration of quantified factors, the river subregions are prioritized by the approximation insertion method and then allocated to UAVs. Moreover, to meet the terminal time constraint, the so-called positive/negative greedy method is proposed to expand or contract waypoints. Finally, the performance of our proposed algorithm is evaluated by simulations on a real river map, and the results verify its good performance in various scenarios.

How to use environmental information to improve the cooperative chasing efficiency of multiple unmanned surface vehicles (multi-USV) is a problem worthy of attention. This article considers an archipelago scenario and proposes an obstacle-assisted chasing framework based on reinforcement learning, in which a multi-USV system chases a smart evader through autonomous environmental awareness and target state prediction. When there are obstacles near the evader, the chasing group can use them to quickly complete the chase. Otherwise, within the range of perception, the group can drive the evader closer to the obstacles or directly cooperate to catch the evader. At the same time, the problem of group credit allocation is also considered, so as to achieve a relatively stable encircling structure with the least number of vehicles. Simulation experiments in the multi-obstacle environment demonstrate the flexibility of the proposed framework. Compared with the traditional force-based and learning-based methods in untrained archipelago environments, the model trained by this framework is proven to have higher efficiency while ensuring real-time performance and generalization. Note to Practitioners —The motivation of this paper is to provide guidance to the USV team in learning how to collaboratively chase and capture target through multi-agent reinforcement learning. Meanwhile, this method can also be applied to other chasing scenarios for vehicles, such as ground robots. This paper proposes a new method that leverages the MA-POCA learning framework and RSA mechanism to enhance cooperation and credit allocation among USVs. Additionally, a novel reward mechanism is designed to encourage USVs to fully utilize the advantages of the surrounding environment and accelerate cooperative chasing. With this method, multiple USVs can effectively pursue intelligent evaders through autonomous environmental perception and target state prediction. To validate the proposed method, we developed a high-fidelity simulation environment using Unity3D, incorporating USV dynamics, irregular obstacles, and sea surface disturbances. Furthermore, our method demonstrates low computational cost, making it suitable for practical USV navigation and control applications in the future.

Objective and Impact Statement . There is a need to develop high-performance and low-cost data augmentation strategies for intelligent skin cancer screening devices that can be deployed in rural or underdeveloped communities. The proposed strategy can not only improve the classification performance of skin lesions but also highlight the potential regions of interest for clinicians’ attention. This strategy can also be implemented in a broad range of clinical disciplines for early screening and automatic diagnosis of many other diseases in low resource settings. Methods . We propose a high-performance data augmentation strategy of search space 10 ¹ , which can be combined with any model through a plug-and-play mode and search for the best argumentation method for a medical database with low resource cost. Results . With EfficientNets as a baseline, the best BACC of HAM10000 is 0.853, outperforming the other published models of “single-model and no-external-database” for ISIC 2018 Lesion Diagnosis Challenge (Task 3). The best average AUC performance on ISIC 2017 achieves 0.909 (±0.015), exceeding most of the ensembling models and those using external datasets. Performance on Derm7pt archives the best BACC of 0.735 (±0.018) ahead of all other related studies. Moreover, the model-based heatmaps generated by Grad-CAM++ verify the accurate selection of lesion features in model judgment, further proving the scientific rationality of model-based diagnosis. Conclusion . The proposed data augmentation strategy greatly reduces the computational cost for clinically intelligent diagnosis of skin lesions. It may also facilitate further research in low-cost, portable, and AI-based mobile devices for skin cancer screening and therapeutic guidance.

This paper focuses on planning a three-dimensional (3-D) path for the autonomous underwater vehicle in ocean environment with complicated static and dynamic obstacles as well as the ocean current. Imitating the principle of natural flow avoiding rocks, we develop a submerged path planning method based on improved interfered fluid dynamical system (IIFDS). In view of the particular feature of ocean environment, the obstacles and ocean current are modeled first. The flow field weight coefficient is then proposed to add the current into the confluence of IIFDS, so that the planned path could make use of the current. In order to obtain the energy-optimal path, the improved genetic algorithm whose mutation operator is modified by grey wolf optimizer is proposed to optimize the repulsive and tangential reaction coefficient, tangential direction coefficient, and flow field weight coefficient. Furthermore, a reverse-avoidance strategy is applied to real-time path planning to avoid dynamic obstacles, combined with IIFDS. Finally, the good performance of our proposed method is verified by simulations in various scenarios.