As part of his thesis, Dario Ziswiler developed a "home sleep lab" with the goal of detecting sleep apnea without visiting a sleep lab.
Dario Ziswiler's work is based on a number of previous bachelor's and master's theses, which were all related to the same topic.
Sleep apnea is a sleep disorder that is manifested by snorting and snoring sounds and breathing pauses of up to 10 seconds. Diagnosis of this disorder is usually done in a sleep lab. However, the many electrodes and cables used in this process can be disturbing and have a negative effect on the patient's sleep.
Therefore, a system should be developed that detects sleep apnea without contact and can be used at home in a familiar environment.
Solution
The measurement system used a 3D-ToF camera to examine volume changes in the patient's chest and abdomen. Based on these volume changes, the three relevant parameters for sleep apnea - respiratory volume, respiratory rate and respiratory failure - were calculated. The calculations were performed in the supine, prone, and lateral lying positions. In addition, the effect of using a blanket on the breathing algorithms was investigated. The code was adapted accordingly in order to be able to perform measurements over a night.
Realization
As part of the realization, the calculations were programmed in Python. The volume of the chest and abdomen was determined by integrating the individual pixel volumes. Various measurements took place to obtain meaningful results.
Results
The results showed that the respiratory volume in supine position could be calculated with an average relative error of 2.63% for inhalation and 0.46% for exhalation compared to the reference system. The average respiratory rate and various breathing stops could also be reliably determined. Similar results were obtained in lateral position, while in prone position the error in respiratory volume was slightly larger. Examination of ceiling effects revealed that pose detection worked well as long as shoulders, elbows, and hands were not covered by the ceiling. However, as soon as the patient was completely covered, correct pose detection was no longer possible. Through experimentation with different blankets, it was determined that a brown polyester blanket provided the best results. The code was prepared accordingly to allow measurements to be taken over an entire night.
What next?
Looking to the future, the well-performing breathing algorithm can be transferred to an embedded system to present a functional overall system.
Further information on the Sensors.ch website
