Research Article
Evaluation of 3-Dimensional In-Body Device Localization Using Machine Learning Algorithms with Reduced Training Data
@INPROCEEDINGS{10.1007/978-3-032-38657-1_5, author={Taiki Egusa and Takumi Kobayashi and Jaakko Hyry and Manato Fujimoto and Shingo Ata and Daisuke Anzai}, title={Evaluation of 3-Dimensional In-Body Device Localization Using Machine Learning Algorithms with Reduced Training Data}, proceedings={Body Area Networks. 20th EAI International Conference, BodyNets 2025, Okayama, Japan, December 2--4, 2025, Proceedings}, proceedings_a={BODYNETS}, year={2026}, month={10}, keywords={Cybernetic Avatars In-body Localization Machine Learning}, doi={10.1007/978-3-032-38657-1_5} }- Taiki Egusa
Takumi Kobayashi
Jaakko Hyry
Manato Fujimoto
Shingo Ata
Daisuke Anzai
Year: 2026
Evaluation of 3-Dimensional In-Body Device Localization Using Machine Learning Algorithms with Reduced Training Data
BODYNETS
Springer
DOI: 10.1007/978-3-032-38657-1_5
Abstract
Recently, active research has been conducted on the development of in-body cybernetic avatars (in-body CA) that can perform actions and operations on behalf of humans within the body with the aim of achieving a society with health longevity outcomes. These avatars are composed of various implantable devices, and it is important to acquire detailed information in vivo to localize device positions. We present a device localization method based on magnetic fields and evaluate the position estimation accuracy of the various machine learning algorithms using a bioequivalent liquid phantom that simulates the human body, to identify which algorithms are suitable for position estimations of in-body devices. In addition, since collecting training data is time-consuming, we also evaluate the relationship between the amount of training data and estimation accuracy, to reduce the required data volume while maintaining an acceptable level of error.