Sleep is fundamental for physical health and good quality of life, and clinicians and researchers have long debated how best to understand it. Manual approaches to sleep classification have been in use for over 40 years, and in 2007, the American Academy of Sleep Medicine (AASM) published a new sleep scoring manual. Over the years, many attempts have been made to introduce and validate machine learning and automated classification techniques in the sleep research field, with the goals of improving consistency and reliability. This thesis explored and assessed the use of automated classification systems with the updated sleep stage definitions and scoring rules using neuro-fuzzy system (NFS) and support vector machine (SVM) methodology. For both the NFS and SVM classification techniques, the overall percent correct was approximately 65%, with sensitivity and specificity rates around 80% and 95%, respectively. The overall Kappa scores, one means for evaluating system reliability, were approximately 0.57 for both the NFS and SVM, indicating moderate agreement that is not accidental. Stage 3 sleep was detected with an 87-89% success rate. The results presented in this thesis show that the use of NFS and SVM methods for classifying sleep stages is possible using the new AASM guidelines. While the current work supports and confirms the use of these classification techniques within the research community, the results did not indicate a significant difference in the accuracy of either approach-nor a difference in one over the other. The results suggest that the important clinical stage 3 (slow wave sleep) can be accurately scored with these classifiers; however, the techniques used here would need more investigation and optimization prior to serious use in clinical applications.
Library of Congress Subject Headings
Sleep--Stages--Classification; Classification--Data processing; Neural networks (Computer science); Support vector machines
Department, Program, or Center
Computer Engineering (KGCOE)
Fehrmann, Elizabeth, "Automated sleep classification using the new sleep stage standards" (2013). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus