Recently, deep learning-based methods have emerged as the preferred approach for ultrasound data analysis. However, these methods often require large-scale annotated datasets for training deep models, which are not readily available in practical scenarios. Additionally, the presence of speckle noise and other imaging artifacts can introduce numerous hard examples for ultrasound data classification. In this paper, drawing inspiration from self-supervised learning techniques, we present a pre-training method based on mask modeling specifically designed for ultrasound data. Our study investigates three different mask modeling strategies: random masking, vertical masking, and horizontal masking. By employing these strategies, our pre-training approach aims to predict the masked portion of the ultrasound images. Notably, our method does not rely on externally labeled data, allowing us to extract representative features without the need for human annotation. Consequently, we can leverage unlabeled datasets for pre-training. Furthermore, to address the challenges posed by hard samples in ultrasound data, we propose a novel hard sample mining strategy. To evaluate the effectiveness of our proposed method, we conduct experiments on two datasets. The experimental results demonstrate that our approach outperforms other state-of-the-art methods in ultrasound image classification. This indicates the superiority of our pre-training method and its ability to extract discriminative features from ultrasound data, even in the presence of hard examples.

Annotated large-scale datasets are crucial for pathology image analysis, yet creating such datasets is challenging. Self-supervised learning (SSL) offers a potential solution to reduce the need for extensive annotations. However, applying SSL methods designed for natural images directly to pathology images may not yield optimal efficiency and performance due to the unique characteristics of pathological images. In this paper, we explore the strengths and weaknesses of two widely used SSL categories, including generation-based SSL and contrastive SSL in pathology tasks. Our findings indicate that pathology images exhibit a high degree of information redundancy, making the direct application of these SSL categories inefficient and suboptimal. To address this, we propose the Rapid Histopathology Pre-training (RHP) framework, which employs a simple yet effective method for training vision Transformers. RHP uses a dual-branch random masking strategy to integrate the characteristics of pathological images with the advantages of both self-supervised schemes. Additionally, an extra multi-stage strategy can enhance model performance while mitigating domain discrepancy issues. This approach enables rapid training on pathological images, yielding superior performance compared to baseline methods. For our experiments, we extracted 4,821,041 multi-scale pathological images from 37,378 pathology slides for pre-training. We evaluated RHP’s performance on four downstream tasks, including patch classification, nucleus segmentation, gland segmentation, and whole slide image classification across eight public datasets. The results demonstrate that our RHP-based visual representation achieved state-of-the-art performance on each dataset with high efficiency, proving to be more robust compared to other SSL methods and ImageNet transfer learning. Our code is publicly released on GitHub (https://github.com/AndrewTal/RHP).

Sound event detection (SED) refers to recognizing the sound events in a continuous audio signal, which has drawn increasing interest during recent decades. The applications of SED seem to be evident in many fields, ranging from surveillance to monitoring applications. Despite the sustainable efforts that have been made, most of the previous attempts are performed on the closed-set, as only fixed and known sound event classes can be employed during the training. In this paper, we present our incremental few-shot SED framework under the open-set settings, as a practical machine listening system should be able to address unknown sound events. Specifically, an explicit learning and calibration-based multi-stage learning framework is utilized to address the challenges of catastrophic forgetting, and aim to achieve a better trade-off between stability and plasticity. To compress the model efficiently, the model prune and self-distillation paradigm are combined used for the model compression, thus our system can be deployed for the resource-limited devices. Our framework can also provide an uncertain estimation for the inference. Lastly, an interactive interface is presented to demonstrate the functions of our system.