Current Issue [Vol. 12, No. 05] [May 2026]

This study proposes a metaverse-based digital twin system architecture for organ-based physiological anomaly detection and multi-class anomaly classification in the field of healthcare. The proposed system architecture was created by combining synthetic scenario generation, six-dimensional feature extraction, LSTM-based prediction model training, a threshold determination mechanism based on organ functions, a semantic inference-based decision-making module, and Unity-based interactive visualization components. In the proposed system, anomaly detection is performed for the left lung, right lung, and heart using heart rate and respiratory data, and the results are classified as normal, warning, lung anomaly, heart anomaly, and critical condition. Experimental evaluation was performed on a synthetic dataset of 24,000 sample data (each containing 20 time steps). The training model achieved 89.39% accuracy, 82.41% Micro-F1 score, 82.00% Macro-F1 score, 80.34% precision, 84.59% recall and 72.17% Exact Match Accuracy in the test dataset. Consequently, the findings indicate that the proposed approach provides a technically balanced and evaluable solution for multi-label physiological monitoring processes.
Keywords: Digital twin, metaverse, LSTM, physiological time-series, multi-label anomaly detection, healthcare monitoring, Unity visualization

Quantum computing utilizes the basic principles of quantum mechanics, involving superposition and entanglement, for executing computational tasks that are intractable for classical computers [1]. The new paradigm uses the unique quantum properties of qubits for processing as well as storing data, providing exponential speedups of certain problems, such as prime factorization, database search, and simulation regarding complex quantum systems [2], [3].