The VinUni-Illinois Smart Health Center (VISHC) research team is delighted to announce the acceptance of six new papers in prestigious AI and Medicine journals, which include IEEE Transactions on Image Processing, IEEE Transactions on Network and Service Management, PLOS ONE, Scientific Reports (Nature), the International Scientific Journal Engineering Applications of Artificial Intelligence, and Cureus Journal of Medical Science. Our research spans a diverse array of AI and Medicine topics, encompassing both foundational studies and various facets of intelligent healthcare applications. Below we are enthusiastic about sharing our latest publications with you. A full list of publications can be found here.

1. Q. Nguyen, H. H. Pham, K. -S. Wong, P. L. Nguyen, T. T. Nguyen and M. N. Do, “FedDCT: Federated Learning of Large Convolutional Neural Networks on Resource Constrained Devices Using Divide and Collaborative Training,” in IEEE Transactions on Network and Service Management (12 September 2023), doi: 10.1109/TNSM.2023.3314066.

Abstract: We introduce FedDCT, a novel distributed learning paradigm that enables the usage of large, high-performance CNNs on resource-limited edge devices. As opposed to traditional FL approaches, which require each client to train the full-size neural network independently during each training round, the proposed FedDCT allows a cluster of several clients to collaboratively train a large deep learning model by dividing it into an ensemble of several small sub-models and train them on multiple devices in parallel while maintaining privacy. In this co-training process, clients from the same cluster can also learn from each other, further improving their ensemble performance. In the aggregation stage, the server takes a weighted average of all the ensemble models trained by all the clusters. FedDCT reduces the memory requirements and allows low-end devices to participate in FL. We empirically conduct extensive experiments on standardized datasets. Experimental results show that FedDCT outperforms a set of current SOTA FL methods with interesting convergence behaviors. View detail

2. K. -N. C. Mac, M. N. Do and M. P. Vo, “Efficient Human Vision Inspired Action Recognition using Adaptive Spatiotemporal Sampling,” in IEEE Transactions on Image Processing (31 August 2023), doi: 10.1109/TIP.2023.3310661.

Abstract: Adaptive sampling that exploits the spatiotemporal redundancy in videos is critical for always-on action recognition on wearable devices with limited computing and battery resources. The commonly used fixed sampling strategy is not context-aware and may under-sample the visual content, and thus adversely impacts both computation efficiency and accuracy. Inspired by the concepts of foveal vision and pre-attentive processing from the human visual perception mechanism, we introduce a novel adaptive spatiotemporal sampling scheme for efficient action recognition. Our system pre-scans the global scene context at low-resolution and decides to skip or request high-resolution features at salient regions for further processing. We validate the system on EPIC-KITCHENS and UCF-101 datasets for action recognition, and show that our proposed approach can greatly speed up inference with a tolerable loss of accuracy compared with those from state-of-the-art baselines. View detail

3. Thuy Dung Nguyen, Tuan Nguyen, Phi Le Nguyen, Hieu H. Pham, Khoa Doan, and Kok-Seng Wong. “Backdoor Attacks and Defenses in Federated Learning: Survey, Challenges and Future Research Directions” – International Scientific Journal Engineering Applications of Artificial Intelligence (Accepted, September 2023 – To appear).

Abstract: Federated learning (FL) is a machine learning (ML) approach that allows the use of distributed data without compromising personal privacy. However, the heterogeneous distribution of data among clients in FL can make it difficult for the orchestration server to validate the integrity of local model updates, making FL vulnerable to various threats, including backdoor attacks. Backdoor attacks involve the insertion of malicious functionality into a targeted model through poisoned updates from malicious clients. These attacks can cause the global model to misbehave on specific inputs while appearing normal in other cases. Backdoor attacks have received significant attention in the literature due to their potential to impact real-world deep-learning applications. However, they have not been thoroughly studied in the context of FL. In this survey, we provide a comprehensive survey of current backdoor attack strategies and defenses in FL, including a comprehensive analysis of different approaches. We also discuss the challenges and potential future directions for attacks and defenses in the context of FL. View detail

4. Tu N. Doan Thu, Quan K. Nguyen, Andrew W. Taylor-Robinson. “Healthcare in Vietnam: Harnessing Artificial Intelligence and Robotics to Improve Patient Care Outcomes”. Cureus Journal of Medical Science 15(9), e45006 (11 September, 2023), doi:10.7759/cureus.45006

Abstract: Healthcare in Vietnam is increasingly utilizing artificial intelligence (AI) and robotics to enhance patient care outcomes. The Vietnamese healthcare sector recognizes the potential of AI and is actively exploring its applications in research and clinical practice. AI technologies, such as text mining and machine learning, can be employed to analyze medical data and improve decision-making processes. Robotics, on the other hand, can support various healthcare tasks, including elderly care, rehabilitation, and surgical interventions. Robotic surgery, specifically, is an innovative form of minimally invasive surgery that aims to improve surgical outcomes and enhance the patient experience. The implementation of AI in emergency and trauma settings is still in its early stages, but there is a growing interest in and recognition of its potential benefits. However, there are challenges that need to be addressed, such as the need for appropriate research and training programs to support the adoption and integration of AI in healthcare. Despite these challenges, healthcare professionals in Vietnam are optimistic about the potential of AI to improve acute care surgery and are open to embracing new digital technologies. The use of AI and robotics in healthcare aligns with the broader goal of improving healthcare systems in low- and middle-income countries, including Vietnam, through technological advancements. Overall, AI can play an important role in assisting prognosis and predictive analysis by integrating vast amounts of data. Moreover, the integration of AI and robotics in healthcare in Vietnam has the potential to enhance patient care outcomes, improve decision-making processes, and support healthcare professionals in their practice. View detail

5. Anh Duy Nguyen, Huy Hieu Pham, Huynh Thanh Trung, Quoc Viet Hung Nguyen, Thao Nguyen Truong, Phi Le Nguyen. “High Accurate and Explainable Multi-Pill Detection Framework with Graph Neural Network-Assisted Multimodal Data Fusion” – PLOS ONE (Accepted, September 2023 – To appear).

Abstract: Due to the significant resemblance in visual appearance, pill misuse is prevalent and has become a critical issue, responsible for one-third of all deaths worldwide. Pill identification, thus, is a crucial concern needed to be investigated thoroughly. Recently, several attempts have been made to exploit deep learning to tackle the pill identification problem. However, most published works consider only single-pill identification and fail to distinguish hard samples with identical appearances. Also, most existing pill image datasets only feature single pill images captured in carefully controlled environments under ideal lighting conditions and clean backgrounds. In this work, we are the first to tackle the multi-pill detection problem in real-world settings, aiming at localizing and identifying pills captured by users in a pill intake. Moreover, we also introduce a multi-pill image dataset taken in unconstrained conditions. To handle hard samples, we propose a novel method for constructing heterogeneous a priori graphs incorporating three forms of inter-pill relationships, including co-occurrence likelihood, relative size, and visual semantic correlation. We then offer a framework for integrating a priori with pills’ visual features to enhance detection accuracy. Our experimental results have proved the robustness, reliability, and explainability of the proposed framework. Experimentally, it outperforms all detection benchmarks in terms of all evaluation metrics. Specifically, our proposed framework improves COCO mAP metrics by 9.4% over Faster R-CNN and 12.0% compared to vanilla YOLOv5. Our study opens up new opportunities for protecting patients from medication errors using an AI-based pill identification solution. View detail

6. Son Hai Nguyen, Phan Kim Thi Vu and Mai Thi Tran. “Absorbance biosensors-based hybrid MoS2 nanosheets for Escherichia coli detection”, in Scientific Reports (Nature), Volume 13, Article number: 10235 (July 2023). doi https://doi.org/10.1038/s41598-023-37395-4.

Abstract: Detecting Escherichia coli is essential in biomedical, environmental, and food safety applications. In this paper, we have developed a simple, rapid, sensitive, and selective E. coli DNA sensor based on the novel hybrid-type MoS2 and (NH4)6Mo7O24 nanosheets. The sensor uses the absorbance measurement to distinguish among the DNA of E. coli, Vibrio proteolyticus, and Bacillus subtilis when implemented in conjunction with NH2-probes. Our experiments showed that the absorbance increased when sensors detected E. coli DNA, whereas it decreased when sensors detected V. proteolyticus and B. subtilis DNA. To the best of authors’ knowledge, there are no reports using the novel hybrid-MoS2 and (NH4) 6Mo7O24 materials for differentiating three types of DNA using cost-effective and rapid absorbance measurements. In addition, the label-free E. coli DNA biosensor exhibited a linear response in the range of 0 fM to 11.65 fM with a limit of detection of 2 fM. The effect of NH2-probes on our sensors’ working performance is also investigated. Our results will facilitate further research in pathogen detection applications, which have not been fully developed yet. View detail