Aim:

             We proposed detecting Sleep Apnea Detection From Single-Lead ECG. The advancement of smart wearables technologies has provided a unique opportunity for sleep and health monitoring. However, wearable technologies rely on accurate and real-time monitoring algorithms.

Abstract:                         

           This paper presents a fully integrated system for the detection and prevention of sleep apnea in infants. Apnea has been one of the leading causes of death worldwide with an approximation of about 200 deaths of premature neonatal infants a year. Currently, for the diagnosis of apnea the patients need to go through overnight sleep study in the laboratory, which is very expensive. The proposed device will be a solution for both monitoring and preventing the condition using accurate readings and also judging and giving suggestions on when the patient needs medical help using cloud and artificial intelligence. And all this done in the respective homes of the patients

Existing System:

          Technologies presented in recent literatures for apnea include monitoring the oxygen levels using sensors and, in some cases, include chest belt, straingauge, etc. These methods are not suitable for neonatal respiratory monitoring because babies cannot wear these devices due to the sensitive nature of their skin. The proposed device uses a spo2 sensor, that represents a very small form factor and consumes very small amount of power which makes it suitable for daily home usage. An algorithm used with the MAX30102 (For Spo2) output signal can make up for the error associated with Spo2 readings with ambient temperature changes. By this the device could help notify the variations in the heart rate of the baby which is caused due to the variations in oxygen levels leading to complications. This also detects the room temperature from which we can make sure it’s apt for the baby.

Proposed System:

            We provide a fair and unbiased comparison between different conventional machine learning and deep learning algorithms in the detection of sleep apnea occurrence from a single-lead ECG. All the experiments are performed on the same dataset and under the same setting to be able to properly evaluate and compare the performances of different algorithms. Unlike most studies that tune their model hyperparameters based on the same data used for final evaluation, we used three sets of data: a training set to train the model parameters, a validation set to find the model optimum hyperparameters, and a test set to evaluate the generalizability of the developed models on unseen data. where the performance of a few deep learning methods was analyzed for the detection of sleep apnea. We also performed a feature importance analysis and demonstrated which ECG features are most effective for the detection of apnea episodes.