Newborn infant breathing remote detection using optical sensor for alarm generation

Details

Supervisors

Legat, Jean-Didier ; Doguet, Pascal

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical

Abstract

Breathing troubles are very common pathologies for newborn children and remain unfortunately the most important causes of death despite information campaigns and researches. Among them, Sudden Infant Death Syndrome, Sleep Apnea and Accidental Suffocation are the most common ones and often lead to the death of the baby if a medical care is not immediately brought. There is then an urgent need in a reliable and efficient monitoring device for the breathing trouble detection and the alarm generation calling for medical care. All the existing detection devices require direct contact with the babies, disturbing or even injuring them, or indirect contact which decreases the efficiency of the detection. Therefore, the company Synergia Medical launched the VitalSyn project for the development of a non-contact reliable and affordable monitoring device. In the scope of this project, this thesis investigated the feasibility of a multipixel IR thermopile sensor for the remote detection of thermal variations in the region of subject's nose. The objectives for this thesis is to validate the choice of the sensor and to determine its performances in the detection of apnea inside breathing signals. During this study, we performed our tests for two different multipixel IR thermopile sensors (Heimann's and Melexis sensors). The algorithm for an efficient detection was divided in two parts. The first part, called the AUTOPERIOD algorithm, was implemented in order to determine if a sample of the signal was a breathing or an apnea by a periodicity detection. This first method was then used in order to determine the maximum acquisition distance of the sensors when placed under the nose or in front of the subject. The second part of the method was called the Apnea Detection Method. This algorithm was designed to use the AUTOPERIOD method over the entire signal using the sliding window technique in order to identify where apnea appears in the signal. This second part was used in order to determine the precision of detection, processing times and delay between apnea and its detection. Results of the different tests showed that the Melexis sensor was the most precise sensor and could perform efficient detections at a distance of 70 cm under subject's nose and 40 cm when placed in front of the face. Tests on apnea detection method showed that detections could be precisely performed for apnea of at least 20 seconds. They also showed processing times far lower than acquisition times introducing no delay. Finally, delays between apnea and its detection for the best tuning of parameters were identified between 4 and 8 seconds. In conclusion, this thesis permitted to show promising results for the use of multipixel IR thermopile sensors for the remote breathing detection.In order to be completely reliable, some improvements and further tests should however be conducted as real-time analysis or the automation of the tuning of some parameters.