The goal of this project is to develop a low cost automated respiratory event scoring system that emulates a particular somnologist's analysis protocol. Respiratory event scoring is critical for diagnosing sleep disordered breathing (SDB). This automated process is a needed tool in sleep medicine since it reduces the tedious visual analysis required by long duration recordings. SDB affects over 18 million Americans; many are undiagnosed because sleep studies are costly and time consuming. Although a number of automated systems have been developed for respiratory event scoring; none have met the reliability criteria required by sleep clinicians. All of these systems are rule-based expert systems which deviate greatly from a sleep expert's scoring strategy. Respiratory event scoring is not a standardized task and almost every clinic has a unique scoring strategy and recording montage. The system to be developed will have a high somnologist/computer agreement for a particular sleep expert. The entire project will address a range of patient age groups with varying SDB type and severity through algorithm adaptation to individual somnologists. Using neural network and fuzzy logic technology, the algorithms will implement the somnologist's criteria and his preferred recording montage.
A large customer base exists and is growing in sleep medicine, particularly in diagnosing SDB. With a successful large validation study this product will be first automated respiratory event scoring system to be accepted by the sleep community. Reliability, low cost, and accessibility for any lab with a MS-Windows or Macintosh system will be the selling points of this system.
