SPEAKERS

Tho Nguyen is currently a Head of the international Unit of IAD, Coordinator & Researcher. He is also a doctor of Preventive Medicine, graduated from Hue University of Medicine and Pharmacy, and is a researcher at the International Research Institute for Artificial Intelligence and Data Science, Dong A University.

He is eager to work on health data that helps him figure out casual relationships among health problems. With his medical knowledge and data analysis skills in machine learning, he believes that more medical research involving machine learning and AI will be conducted in the future.

Tho Nguyen has contributions in a Chapter of a Book: "Artificial Intelligence for Smart manufacturing" and a proceeding in an international conference: "ISSAT International Conference on Data Science in Business, Finance and Industry" (DSBFI 2023).

Research interests: 

• Medical research involving machine learning and AI

Title: From Blackbox to Explainable Artificial Intelligence for Healthcare 5.0

Abstract:

Artificial Intelligence (AI) has revolutionized the capabilities of diagnosis and prediction in healthcare, aiding clinical experts in the decision-making process. However, AI models require more explainability about decisions made within black boxes, raising concerns about the transparent accuracy of outcomes. The emergence of Explainable Artificial Intelligence (XAI) has narrowed the gap between the need for explainability in clinical decisions and the remarkable performance of AI algorithms. XAI elucidates the features that influence decisions driven by AI algorithms; this approach supports clinical experts with a profound understanding of the origins of decisions and assists them in making intelligent choices.

Healthcare 5.0 emphasizes the essential well-being of humans through real-time monitoring, accurate assessment, and diagnosis facilitated by AI while upholding patient data privacy and robust security. The Federated Learning (FL) technique involves training an AI model across various distributed devices (local servers) by exchanging weight parameters of models with a central global server. XAI models trained using FL tend to achieve higher accuracy with interpretable models, low computation time, data privacy, and robust security. To provide a comprehensive illustration of training XAI on FL, in this presentation, we showcase an artificial intelligence experiment in fall detection using smart wearables, which holds the potential for safeguarding the lives of the elderly against fall-induced accidents and protecting workers in manufacturing settings. Furthermore, the outcomes of pneumonia detection from chest X-ray images using the XAI model are interpreted to comprehend the operations of the black box.

Keywords: Explainable Artificial Intelligence, black box, Federated learning, Healthcare 5.0