The PerfectPillow

by Reha Matai

Coauthors: Thank you to Dr. Sharief Taraman, Dr. Anthony Chang, and Vanessa Rohrer for this amazing opportunity to participate in the CHOC MI3 Summer Internship, as well as your support and guidance in the development of this abstract.

Medical Devices & Digital Health

Obstructive sleep apnea (OSA), which is a sleep-related respiratory disorder, is becoming more common in children. Although, approximately 80-90 percent of moderate and severe cases in children go undiagnosed. OSA is caused by upper airway resistance when a child is asleep, leading to an interruption of airflow shown from heavy breathing, restless sleep, and gasps while snoring. The current gold-standard approach for OSA detection is polysomnography (PSG), which examines a patient’s sleep signals overnight while the patient is connected to various sensors. However, this method has several disadvantages as it doesn’t account for those who can’t sleep well in a hospital environment or are disturbed by the attached electrodes, resulting in inaccurate readings.

The PerfectPillow is a tech-enabled headrest cushion that resembles and serves as a pillow, meant to be placed underneath the patient’s head to closely track their vital signs, eliminating the need to connect electrodes. To track heart and respiratory rate, a pneumatic sensor is integrated into the pillow to monitor air-pressure fluctuations, which are usually caused by tiny tremors from heartbeats, uneven breaths, or constant turning on the bed. More specifically, the sensor will measure these through ballistocardiography and record these ballistic forces coming from the heart. Additionally, there is an accelerometer embedded in the pillow that collects data on how quickly and how much the patient moves throughout the night, thus checking for restless sleep. To detect snoring and difficulty breathing, a sound sensor actively records any noise heard throughout the night by identifying specific snoring audio signals and periods of cessation of breathing. This feature also helps inform the patient if they’re dealing with any other forms of loud interruptions, possibly contributing to their sleep apnea. As for the PerfectPillow’s physical design, it’s made of an attractive airy lightweight memory foam material and is offered in various styles, colors, and sizes as per the patient’s preference. The pillow’s integrated features and design allows for the patient to sleep as they normally would, alleviating the discomfort and disturbance that comes from attaching multiple devices.

The data of the patient’s heart rate, respiratory rate, movement, and noise levels from the PerfectPillow will automatically sync to the SleepMate app via Bluetooth. This app provides an interactive way of interpreting the data collected with visualization of trends over time. A machine learning algorithm is built into the app to perform a detailed analysis of this information and detect sleep anomalies. To identify complete OSA events, a window overlapping method is used to recognize the start and end positions of the OSA event. This deep neural network model will be trained based on the data collected from known OSA patients and use LSTM (long short-term memory) to maintain internal memory and perform feedback connections to learn from a sequence of inputs. In this way, the model helps diagnose OSA conditions with confidence. Using machine learning to analyze the collected data, the contactless PerfectPillow paired with the SleepMate app provides a promising alternative to increase patient satisfaction and most importantly, comfort.