Hysteresis is a unique type of dynamic, which contains an important property, rate-independent memory. In addition to other memory-related studies such as time delay neural networks, recurrent networks, and reinforcement learning, rate-independent memory deserves further attention owing to its potential applications. In this paper, we attempt to define hysteretic memory (rate independent memory) and examine whether or not it could be modeled in neural networks. Our analysis results demonstrate that other memory-related mechanisms are not hysteresis systems. A novel neural cell, referred to herein as the propulsive neural unit, is then proposed. The proposed cell is based on a notion related the submemory pool, which accumulates the stimulus and ultimately assists neural networks to achieve model hysteresis. In addition to training by backpropagation, a combination of such cells can simulate given hysteresis trajectories.
Diabetes mellitus (DM) foot ulcer is a chronic wound and is highly related to the mortality and morbidity of infection, and might induce sepsis and foot amputation, especially during the isolation stage of the COVID-19 pandemic. Visual observation when changing dressings is the most common and traditional method of detecting wound healing. The formation of granulation tissues plays an important role in wound healing. In the complex pathophysiology of excess and unhealthy granulation induced by infection, oxygen supply may explain the wound healing process in DM patients with multiple complicated wounds. Thus, advanced and useful tools to observe the condition of wound healing are very important for DM patients with extremities ulcers. For this purpose, we developed an artificial intelligence (AI) detection model to identify the growth of granulation tissue of the wound bed. We recruited 100 patients to provide 219 images of wounds at different healing stages from 2 hospitals. This was performed to understand the wound images of inconsistent size, and to allow self-inspection on mobile devices, having limited computing resources. We segmented those images into 32 × 32 blocks and used a reduced ResNet-18 model to test them individually. Furthermore, we conducted a learning method of active learning to improve the efficiency of model training. Experimental results reveal that our model can identify the region of granulation tissue with an Intersection-over-Union (IOU) rate higher than 0.5 compared to the ground truth. Multiple cross-repetitive validations also confirm that the detection results of our model may serve as an auxiliary indicator for assessing the progress of wound healing. The preliminary findings may help to identify the granulation tissue of patients with DM foot ulcer, which may lead to better long-term home care during the COVID-19 pandemic. The current limit of our model is an IOU of about 0.6. If more actual data are available, the IOU is expected to improve. We can continue to use the currently established active learning process for subsequent training.
Cloud computing is an emerging technology developed for providing various computing and storage services over the Internet. In this paper, we proposed a privacy-preserving cloud-aware scenario for compressive multimedia applications, including multimedia compression, adaptation, editing/manipulation, enhancement, retrieval, and recognition. In the proposed framework, we investigate the applicability of our/existing compressive sensing (CS)-based multimedia compression and securely compressive multimedia “trans-sensing” techniques based on sparse coding for securely delivering compressively sensed multimedia data over a cloud-aware scenario. Moreover, we also investigate the applicability of our/existing sparse coding-based frameworks for several multimedia applications by leveraging the strong capability of a media cloud. More specifically, to consider several fundamental challenges for multimedia cloud computing, such as security and network/device heterogeneities, we investigate the applications of CS and sparse coding techniques in multimedia delivery and applications. As a result, we can build a unified cloud-aware framework for privacy-preserving multimedia applications via sparse coding.
13 January 2010 in Annals of Biomedical Engineering
Seam carving is a content-aware image processing approach that has been successfully applied to resizing or removing objects from digital images. Seam-carved images are hard to identify from the original image, making this detection method an important and attractive research topic. Existing methods are based on steganography attacks or statistical features, and their trained and tested models are all constructed using support vector machines (SVMs). The trained models of these methods take a long time to build from a limited number of images. This paper presents a new patch-based sobel operator (PSO) method using SVM and extreme learning machines (ELMs) based on the patch analysis method with square-based features. The ELM adopts five types of neurons to obtain five prediction results, which are then combined using the jury voting scheme in order to improve the accuracy. In addition, combinatorial fusion is first used to choose two of five prediction results according to the diversity rank/score graph, and these two results are then combined to make the final decision. The PSO method in the ELM with jury voting and combinatorial fusion achieves accuracies of 70.60-98.80% and 95.58-99.36% for 10-50% seam-carved and seam-insertion images, respectively. Additionally, for the PSO method, the ELM required only 1.2% of the training time required by the SVM. In conclusion, the PSO method in the ELM is useful for constructing trained models from a large number of images.