A Transformer-Based Semantic Approach to Internet of Medical Things Applications: Integrating Bidirectional Encoder Representations from Transformers Topic Modeling with Dimensionality Reduction

Abstract

This study presents a comprehensive analysis of the prevailing themes and technological advancements within the Internet of Medical Things (IoMT) literature over the past decade (2016–2025). Given its potential to fundamentally transform healthcare delivery, the IoMT represents a rapidly evolving and critical research domain. This analysis aims to identify future research trajectories and pinpoint pivotal challenges within biomedical engineering. Approximately 2000 articles were selected from the Web of Science database using the keyword "Internet of Medical Things." The derived text data was analyzed through the implementation of topic modeling, leveraging the Bidirectional Encoder Representations from Transformers (BERT) Topic method within a Python environment on Google Colab. The findings are presented as document and topic maps, visualized using topic probability distributions and dimensionality reduction techniques such as Uniform Manifold Approximation and Projection (UMAP)/t-Distributed Stochastic Neighbor Embedding (t-SNE). The BERT Topic analysis reveals that IoMT research is predominantly focused on enhancing security and privacy through edge computing and blockchain-based approaches. Furthermore, federated learning and deep learning algorithms demonstrate significant potential for specific medical applications, such as heart and mental disease prediction and COVID-19 management. The results highlight that data security, privacy, and reliability remain the three core themes of the IoMT ecosystem. Ultimately, this study demonstrates that cybersecurity and privacy vulnerabilities constitute the primary barriers to IoMT adoption, suggesting that future research must prioritize the integration of these protective mechanisms within real-time, resource-constrained devices.

Cite this article as: H. Kuduz, G. Tonguç and K. K. Çevik, "A transformer-based semantic approach to internet of medical things applications: integrating bidirectional encoder representations from transformers topic modeling with dimensionality reduction," Electrica, 26, 0060 2026. doi: 10.5152/electrica.2026.26060.