Development of Symbolic Signal Processing and Transformer Models for Predicting Respiratory System Mechanics in Mechanical Ventilation
9
Issued Date
2023-01-01
Resource Type
Scopus ID
2-s2.0-85179550320
Journal Title
BMEiCON 2023 - 15th Biomedical Engineering International Conference
Rights Holder(s)
SCOPUS
Bibliographic Citation
BMEiCON 2023 - 15th Biomedical Engineering International Conference (2023)
Suggested Citation
Junwei Y., Numthavaj P., Pattanateepapon A., Puttanawarut C., Junhasavasdikul D. Development of Symbolic Signal Processing and Transformer Models for Predicting Respiratory System Mechanics in Mechanical Ventilation. BMEiCON 2023 - 15th Biomedical Engineering International Conference (2023). doi:10.1109/BMEiCON60347.2023.10322095 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/91566
Title
Development of Symbolic Signal Processing and Transformer Models for Predicting Respiratory System Mechanics in Mechanical Ventilation
Author's Affiliation
Other Contributor(s)
Abstract
This paper focuses on the assessment of respiratory mechanics, i.e., compliance (C) and resistance (R) on the analysis of respiratory signals. Inspired by the growing use and success of the transformer model in fields such as natural language processing, image recognition, and signal analysis, we have devised an innovative method that leverages automatic feature extraction via transformers to predict C and R. While the use of transformers in respiratory signals has not been widely studied yet, we demonstrate their efficacy for extracting relevant features from respiratory signals in this paper. As transformers require a lot of memory, we have developed a symbolic approach to process the signal, which significantly reduces the size of input data and results in a more compact model. Our experimental findings showed that the proposed algorithm achieved mean absolute errors of 6.91 mL/cmH2O and 3.01 cmH2O.s/L, as well as mean absolute percentage errors of 15% and 20.6% when determining respiratory C and R respectively. These results demonstrated the potential of the proposed method for developing a new generation of ventilation monitoring techniques that could enhance the care given to specific intensive care unit patients.