G5 Doctoral dissertation (article)
Automatic sleep scoring based on multi-modality polysomnography data (2020)


Yan, Rui (2020). Automatic sleep scoring based on multi-modality polysomnography data. JYU dissertations, 298. Jyväskylä: Jyväskylän yliopisto. http://urn.fi/URN:ISBN:978-951-39-8329-1


JYU authors or editors


Publication details

All authors or editors: Yan, Rui

eISBN: 978-951-39-8329-1

Journal or series: JYU dissertations

eISSN: 2489-9003

Publication year: 2020

Number in series: 298

Number of pages in the book: 1 verkkoaineisto (60 sivua, 49 sivua useina numerointijaksoina, 5 numeroimatonta sivua)

Publisher: Jyväskylän yliopisto

Place of Publication: Jyväskylä

Publication country: Finland

Publication language: English

Persistent website address: http://urn.fi/URN:ISBN:978-951-39-8329-1

Open Access: Publication published in an open access channel


Abstract

Over the past decades, probably due to our hectic lifestyle in modern society, complaints about sleep problems have increased dramatically, affecting a large part of the world’s population. The polysomnography (PSG) test is a common tool for diagnosing sleep problems, but the scoring of PSG recordings is an essential but time-consuming process. Therefore, automatic sleep scoring becomes crucial and urgent to settle the growing unmet needs in sleep research. This thesis extends the previous research on automatic sleep scoring from two aspects. One is to extensively explore signal modalities and feature types related to automatic sleep scoring. This exploratory work obtains the optimal signal fusion and feature set for automatic sleep scoring, and further clarifies the contribution of signals and features to the discrimination of sleep stages. Our results demonstrate that diverse features and signal modalities are coordinative and complementary, which benefits the improvement of classification accuracy. The other one is to develop automatic sleep scoring tools that can accommodate different datasets and sample populations without adjusting model structure and parameters across tasks. Experimental results show that the joint analysis of multiple signals can improve the stability, robustness and generalizability of the proposed models. Model performance has been verified on multiple public datasets, demonstrating good model transferability between different datasets and diverse disease populations. In summary, this research finding will advance the understanding of underlying mechanism during automatic sleep scoring and clarify the association between manual scoring criteria and automatic scoring methods. The joint analysis of multiple signals enhances model versatility, which inspires the construction of cross-model in the field of automatic sleep scoring. Moreover, the proposed automatic sleep scoring methods can be integrated with diverse PSG systems, thereby facilitating sleep monitoring in clinical or routine care.


Keywords: brain research; sleep; sleep disorders; signal analysis; signal processing; machine learning

Free keywords: automatic sleep scoring; polysomnography; multi-modality analysis; deap learning; machine learning


Contributing organizations


Ministry reporting: Yes


Last updated on 2020-16-11 at 16:36