Learning Sleep Quality from Daily Logs

pp. 2421-2429, 2019.

Cited by: 1|Bibtex|Views23|Links
EI
Keywords:
data imputation insomnia interpretability precision psychiatry ranking modelMore(1+)
Weibo:
This study, base on real-world logs gathered from insomnia sufferers over a 6-week period, demonstrates how sleep and activity data collected from smart bands can be analyzed to estimate sleep quality

Abstract:

Precision psychiatry is a new research field that uses advanced data mining over a wide range of neural, behavioral, psychological, and physiological data sources for classification of mental health conditions. This study presents a computational framework for predicting sleep efficiency of insomnia sufferers. A smart band experiment is c...More

Code:

Data:

0
Introduction
  • Insomnia is a common psychiatric illness that can decrease quality of life.1 Approximately 30 percent of people from different countries report to suffer from one or more of the symptoms of insomnia [26].
  • Insomnia is a common psychiatric illness that can decrease quality of life.1.
  • 30 percent of people from different countries report to suffer from one or more of the symptoms of insomnia [26].
  • Insomnia sufferers may each exhibit different behavioral characteristics even though their exposed symptoms might be similar.
  • Individual variability, such as the genetic information, neural circuits, individual characteristics, medical codes and electronic health records (EHR) are carefully considered to collectively arrive at a diagnosis, treatment plan, and prediction of prognosis [10, 15].
  • Its premise is that jointly analyzing heterogeneous data sources can yield more accurate classification of major psychiatric illnesses than manual classification like the Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD) [8]
Highlights
  • Insomnia is a common psychiatric illness that can decrease quality of life.1 Approximately 30 percent of people from different countries report to suffer from one or more of the symptoms of insomnia [26]
  • Since our study focus is insomnia, we screened recruits based on the test result of the Insomnia Severity Index (ISI), which is a brief instrument designed to assess the severity of both nighttime and daytime components of insomnia
  • We can find that our Pairwise Learning-based Ranking Generation (PLRG) leads to the highest precision scores, especially when K ≤ 5
  • Such results demonstrate that PLRG is able to more accurately detect users with higher insomnia than competing methods
  • This study, base on real-world logs gathered from insomnia sufferers over a 6-week period, demonstrates how sleep and activity data collected from smart bands can be analyzed to estimate sleep quality
  • Our pairwise learning-based ranking generation model, called PLRG, could rank individuals who will be at high risk of insomnia in the sleep, based on representation learning of sequential and interaction features
Methods
  • Methods of missing data imputation

    Average† Imp-GAIN†

    0.00152 0.00153 0.00162 0.00170 0.00162 0.00162 0.00139 0.00138 model with Imp-GAIN can improve MSE around 9%-16%.4 the authors need to emphasize that while the model displayed good performance, it is capable of interpreting the underlying causes affecting the prediction results: linear, KNN, LASSO regressors are interpretable but they showed lower performance.

    Interpreting user traits based on attention mechanism.
  • The attention mechanism on Phase 2 is to see what step-size s would be the most explainable in predicting the target night’s sleep efficiency, and the attention mechanism on Phase 1 is to see which days are more explainable within the specified s.
  • When observing UserId 8 to predict sleep efficiency values on May 28 and 29 (see Fig. 4(a)(b), the most explainable s were 7 on Phase 2, and the most explainable days were May 25.
  • UserId 40 had the different most interpretable s=5 and corresponding attention day May 26 from UserId 8 (see Fig. 4(c))
Results
  • The authors draw several insights from Fig. 6.
  • The authors can find that the PLRG leads to the highest precision scores, especially when K ≤ 5.
  • Such results demonstrate that PLRG is able to more accurately detect users with higher insomnia than competing methods.
  • The two models with higher precision scores are LSTM-B and DF-XGB, which corresponds to two main components in PLRG.
Conclusion
  • CONCLUSION AND DISCUSSION

    Mental well-being is fundamental to human health. Heterogeneous data sources that are widely becoming available can make a huge impact in psychiatry.
  • This study, base on real-world logs gathered from insomnia sufferers over a 6-week period, demonstrates how sleep and activity data collected from smart bands can be analyzed to estimate sleep quality.
  • In this process, the authors notice that missing data handling becomes a key challenge and propose to impute data via an improved generative adversarial networks, called Imp-GAIN.
  • These models and prediction outcomes have been reviewed by a psychiatrist for a practical use and plan to be used at a hospital for a trial
Summary
  • Introduction:

    Insomnia is a common psychiatric illness that can decrease quality of life.1 Approximately 30 percent of people from different countries report to suffer from one or more of the symptoms of insomnia [26].
  • Insomnia is a common psychiatric illness that can decrease quality of life.1.
  • 30 percent of people from different countries report to suffer from one or more of the symptoms of insomnia [26].
  • Insomnia sufferers may each exhibit different behavioral characteristics even though their exposed symptoms might be similar.
  • Individual variability, such as the genetic information, neural circuits, individual characteristics, medical codes and electronic health records (EHR) are carefully considered to collectively arrive at a diagnosis, treatment plan, and prediction of prognosis [10, 15].
  • Its premise is that jointly analyzing heterogeneous data sources can yield more accurate classification of major psychiatric illnesses than manual classification like the Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD) [8]
  • Objectives:

    The authors' goal is to find f that minimizes the mean absolute error (MAE): minf (i, j)∈M /∥M ∥.
  • The authors aim to make the feature representation capture the two user behaviors
  • Methods:

    Methods of missing data imputation

    Average† Imp-GAIN†

    0.00152 0.00153 0.00162 0.00170 0.00162 0.00162 0.00139 0.00138 model with Imp-GAIN can improve MSE around 9%-16%.4 the authors need to emphasize that while the model displayed good performance, it is capable of interpreting the underlying causes affecting the prediction results: linear, KNN, LASSO regressors are interpretable but they showed lower performance.

    Interpreting user traits based on attention mechanism.
  • The attention mechanism on Phase 2 is to see what step-size s would be the most explainable in predicting the target night’s sleep efficiency, and the attention mechanism on Phase 1 is to see which days are more explainable within the specified s.
  • When observing UserId 8 to predict sleep efficiency values on May 28 and 29 (see Fig. 4(a)(b), the most explainable s were 7 on Phase 2, and the most explainable days were May 25.
  • UserId 40 had the different most interpretable s=5 and corresponding attention day May 26 from UserId 8 (see Fig. 4(c))
  • Results:

    The authors draw several insights from Fig. 6.
  • The authors can find that the PLRG leads to the highest precision scores, especially when K ≤ 5.
  • Such results demonstrate that PLRG is able to more accurately detect users with higher insomnia than competing methods.
  • The two models with higher precision scores are LSTM-B and DF-XGB, which corresponds to two main components in PLRG.
  • Conclusion:

    CONCLUSION AND DISCUSSION

    Mental well-being is fundamental to human health. Heterogeneous data sources that are widely becoming available can make a huge impact in psychiatry.
  • This study, base on real-world logs gathered from insomnia sufferers over a 6-week period, demonstrates how sleep and activity data collected from smart bands can be analyzed to estimate sleep quality.
  • In this process, the authors notice that missing data handling becomes a key challenge and propose to impute data via an improved generative adversarial networks, called Imp-GAIN.
  • These models and prediction outcomes have been reviewed by a psychiatrist for a practical use and plan to be used at a hospital for a trial
Tables
  • Table1: List of data gathered from the smart bands
  • Table2: Results in MAE by various imputation methods
  • Table3: List of notations used in predicting sleep efficiency
  • Table4: Results in MSE by various prediction models
  • Table5: Rank statistics among s using 294 test chunks
Download tables as Excel
Funding
  • This research was supported by Basic Science Research Program (No NRF-2017R1E1A1A01076400) and Next-Generation Information Computing Development Program (No NRF-2017M3C4A7063570) through the National Research Foundation of Korea funded by the Ministry of Science and ICT in Korea
  • This work was also supported by Ministry of Science and Technology (MOST) Taiwan with grants 108-2636-E-006-002 (MOST Young Scholar Fellowship Program) and 107-2218-E-006-040, and supported by Academia Sinica Thematic Research Program with grant AS-107-TP-M05
Reference
  • Rajendra Acharya, Oliver Faust, N Kannathal, TjiLeng Chua, and Swamy Laxminarayan. 2005. Non-linear analysis of EEG signals at various sleep stages. Elsevier Computer Methods and Programs in Biomedicine 80, 1 (2005), 37–45.
    Google ScholarLocate open access versionFindings
  • Bo An, Haipeng Chen, Noseong Park, and VS Subrahmanian. 2016. MAP: Frequency-based maximization of airline profits based on an ensemble forecasting approach. In proc. of the ACM SIGKDD.
    Google ScholarLocate open access versionFindings
  • Jutta Backhaus, Klaus Junghanns, Andreas Broocks, Dieter Riemann, and Fritz Hohagen. 2002. Test–retest reliability and validity of the Pittsburgh Sleep Quality Index in primary insomnia. Journal of Psychosomatic Research 53, 3 (2002).
    Google ScholarLocate open access versionFindings
  • Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 201Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473 (2014).
    Findings
  • Tessa F Blanken, Jeroen S Benjamins, Denny Borsboom, Jeroen K Vermunt, Casey Paquola, Jennifer Ramautar, Kim Dekker, Diederick Stoffers, Rick Wassing, Yishul Wei, et al. 2019. Insomnia disorder subtypes derived from life history and traits of affect and personality. Elsevier Lancet Psychiatry (2019).
    Google ScholarFindings
  • Chris J.C. Burges. 2010. From RankNet to LambdaRank to LambdaMART: An Overview. Technical Report, Microsoft Research (2010).
    Google ScholarFindings
  • Daniel J Buysse, Charles F Reynolds III, Timothy H Monk, Susan R Berman, and David J Kupfer. 1989. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Research 28, 2 (1989), 193–213.
    Google ScholarLocate open access versionFindings
  • Danilo Bzdok and Andreas Meyer-Lindenberg. 2017. Machine learning for precision psychiatry: Opportunites and challenges. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging (2017).
    Google ScholarLocate open access versionFindings
  • Tianqi Chen and Carlos Guestrin. 2016. XGBoost: A Scalable Tree Boosting System. In proc. of the ACM SIGKDD.
    Google ScholarLocate open access versionFindings
  • Edward Choi, Mohammad Taha Bahadori, Elizabeth Searles, Catherine Coffey, Michael Thompson, James Bost, Javier Tejedor-Sojo, and Jimeng Sun. 2016. Multilayer representation learning for medical concepts. In proc. of the ACM SIGKDD.
    Google ScholarLocate open access versionFindings
  • Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua Kulas, Andy Schuetz, and Walter Stewart. 2016. Retain: An interpretable predictive model for healthcare using reverse time attention mechanism. In Advances in Neural Information Processing Systems. 3504–3512.
    Google ScholarLocate open access versionFindings
  • Hui Ding, Goce Trajcevski, Peter Scheuermann, Xiaoyue Wang, and Eamonn Keogh. 2008. Querying and mining of time series data: experimental comparison of representations and distance measures. In proc. of the VLDB.
    Google ScholarLocate open access versionFindings
  • David F Dinges, Frances Pack, Katherine Williams, Kelly A Gillen, John W Powell, Geoffrey E Ott, Caitlin Aptowicz, and Allan I Pack. 1997. Cumulative sleepiness, mood disturbance, and psychomotor vigilance performance decrements during a week of sleep restricted to 4–5 hours per night. Sleep 20, 4 (1997), 267–277.
    Google ScholarLocate open access versionFindings
  • Kelly R Evenson, Michelle M Goto, and Robert D Furberg. 2015. Systematic review of the validity and reliability of consumer-wearable activity trackers. IJBNPA 12, 1 (2015), 159.
    Google ScholarLocate open access versionFindings
  • Brisa S Fernandes, Leanne M Williams, Johann Steiner, Marion Leboyer, André F Carvalho, and Michael Berk. 2017. The new field of ‘precision psychiatry’. BMC medicine 15, 1 (2017), 80.
    Google ScholarLocate open access versionFindings
  • Nancy L Galambos, Andrea L Dalton, and Jennifer L Maggs. 2009. Losing sleep over it: Daily variation in sleep quantity and quality in Canadian students’ first semester of university. Journal of research on adolescence 19, 4 (2009), 741–761.
    Google ScholarLocate open access versionFindings
  • Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David WardeFarley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative Adversarial Nets. In proc. of the NIPS.
    Google ScholarLocate open access versionFindings
  • Gregory Gutin and Abraham P. Punnen. 2007. The Traveling Salesman Problem and Its Variations. Combinatorial Optimization (2007).
    Google ScholarLocate open access versionFindings
  • J Heikenfeld, A Jajack, J Rogers, P Gutruf, L Tian, Tingrui Pan, R Li, M Khine, J Kim, and J Wang. 2018. Wearable sensors: modalities, challenges, and prospects. Lab on a Chip 18, 2 (2018), 217–248.
    Google ScholarFindings
  • Wahyu Hidayat, Toufan D Tambunan, and Reza Budiawan. 2018. Empowering Wearable Sensor Generated Data to Predict Changes in Individual’s Sleep Quality. In proc. of the IEEE ICoICT.
    Google ScholarLocate open access versionFindings
  • Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural computation 9, 8 (1997), 1735–1780.
    Google ScholarLocate open access versionFindings
  • David A. Kalmbach, J. Todd Arnedt, Vivek Pillai, and Christopher L. Drake. 2016. Identifying At-Risk Individuals for Insomnia Using the Ford Insomnia Response to Stress Test. Sleep 39, 2 (2016), 449–456.
    Google ScholarLocate open access versionFindings
  • Uri Kartoun, Rahul Aggarwal, Andrew Beam, Jennifer K. Pai, Arnaub Chatterjee, Timothy P. Fitzgerald, Isaac Kohane, and Stanley Shaw. 2018. Development of an Algorithm to Identify Patients with Physician-Documented Insomnia. Scientific Reports 8 (2018), Article Number: 7862.
    Google ScholarLocate open access versionFindings
  • Vahid Kazemi and Josephine Sullivan. 2014. One millisecond face alignment with an ensemble of regression trees. In proc. of the IEEE CVPR.
    Google ScholarLocate open access versionFindings
  • Minh-Thang Luong, Hieu Pham, and Christopher D Manning. 2015. Effective approaches to attention-based neural machine translation. arXiv preprint arXiv:1508.04025 (2015).
    Findings
  • Hannah Morphy, Kate M Dunn, Martyn Lewis, Helen F Boardman, and Peter R Croft. 2007. Epidemiology of insomnia: a longitudinal study in a UK population. Sleep 30, 3 (2007), 274–280.
    Google ScholarLocate open access versionFindings
  • Sarun Paisarnsrisomsuk, Michael Sokolovsky, Francisco Guerrero, Carolina Ruiz, and Sergio A. Alvarez. 2018. Deep Sleep: Convolutional Neural Networks for Predictive Modeling of Human Sleep Time-Signals. In KDD Deep Learning Day.
    Google ScholarFindings
  • Aarti Sathyanarayana, Jaideep Srivastava, and Luis Fernandez-Luque. 2017. The science of sweet dreams: predicting sleep efficiency from wearable device data. Computer 50, 3 (2017), 30–38.
    Google ScholarLocate open access versionFindings
  • Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems. 5998–6008.
    Google ScholarLocate open access versionFindings
  • Jinsung Yoon, James Jordon, and Mihaela van der Schaar. 2018. GAIN: Missing Data Imputation using Generative Adversarial Nets. In proc. of the ICML.
    Google ScholarLocate open access versionFindings
  • Zhi-Hua Zhou and Ji Feng. 2017. Deep Forest: Towards An Alternative to Deep Neural Networks. In proc. of the IJCAI.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments