Sleep apnea is a common respiratory disorder during sleep. It is characterized by pauses in breathing or shallow breathing during sleep for longer than 10 seconds. Except the fact that not having a proper sleep and being rested for the next day, in some cases the apnea period (not breathing interval) may last more than 30 seconds and this situation can even be fatal. 14% of men and 5% of women suffer from Obstructive Sleep Apnea (OSA) in United States. Patients may experience apnea for more than 300 times in a single night sleep. Polysomnography (PSG) is a multi-parametric recording of biophysiological changes, containing EEG, ECG, SpO2, Nasal Airflow signals, performed during overnight sleep. In this study, a fully automatic apnea detection algorithm is developed and an early warning system is proposed to predict OSA episodes by extracting time-series features of OSA periods and regular respiration using nasal airflow signal. Extracted features are then reduced to improve the performance of the prediction. Support vector machines (SVM), one of the commonly used classification algorithms in medical applications, is implemented for learning and prediction of the OSA episodes. The results show that OSA episodes are predicted with 87.6% of accuracy and 91.3% of sensitivity, 30 seconds before patient faces apnea. By this approach, apnea related health risks can be minimized by foreknowledge.

Objective. In this study, we conducted a comparative analysis of deep convolutional neural network (CNN) models in predicting obstructive sleep apnea (OSA) using electrocardiograms. Unlike other studies in the literature, this study automatically extracts time-frequency features by using CNNs instead of manual feature extraction from ECG recordings. Approach. The proposed model generates scalogram and spectrogram representations by transforming preprocessed 30 s ECG segments from time domain to the frequency domain using continuous wavelet transform and short time Fourier transform, respectively. We examined AlexNet, GoogleNet and ResNet18 models in predicting OSA events. The effect of transfer learning on success is also investigated. Based on the observed results, we proposed a new model that is found more effective in estimation. In total, 152 ECG recordings were included in the study for training and evaluation of the models. Main results. The prediction using scalograms immediately 30 s before potential OSA onsets gave the best performance with 82.30% accuracy, 83.22% sensitivity, 82.27% specificity and 82.95% positive predictive value. The prediction using spectrograms also achieved up to 80.13% accuracy and 81.99% sensitivity on prediction. Per-recording classification suggested considerable results with 91.93% accuracy for prediction of OSA events. Significance. Time-frequency deep features of scalograms and spectrograms of ECG segments prior to OSA events provided reliable information about the possible events in the future. The proposed CNN model can be used as a good indicator to accurately predict OSA events using ECG recordings.

In this study, we conducted a comparative analysis of deep convolutional neural network (CNN) models in predicting Obstructive Sleep Apnea (OSA) using electrocardiograms. Unlike other studies in the literature, this study automatically extracts time-frequency features by using CNNs instead of manual feature extraction from ECG recordings. For this purpose, the proposed model generates scalogram and spectrogram representations by transforming preprocessed 30-sec ECG segments from time domain to the frequency domain using Continuous Wavelet Transform (CWT) and Short Time Fourier transform (STFT), respectively. We examined AlexNet, GoogleNet and ResNet18 models in predicting OSA events. The effect of transfer learning on success is also investigated. Based on the observed results, we proposed a new model that is found more effective in estimation. In total, 152 ECG recordings were included in the study for training and evaluation of the models. The prediction using scalograms immediately 30 seconds before potential OSA onsets gave the best performance with 82.30% accuracy, 83.22% sensitivity, 82.27% specificity and 82.95% positive predictive value. On the other hand, the prediction using spectrograms also provided up to 80.13% accuracy and 81.99% sensitivity on prediction. The results show that the proposed CNN model can be used as a good indicator to accurately predict OSA events using ECG signals.

High resolution medical images obtained by different imaging modalities stored in PACS needs higher storage space and bandwidth because of requiring much space in memory. In this sense, compression of medical images is important for efficient use of database. The main purpose of image compression is to reduce the number of bits representing the image while preserving the image quality and the intensity level of the pixels as much as possible depending on grayscale or RGB image [1]. Since medical images also contain diagnostic information about a disease or an artifact, less or no loss of detail in terms of quality is desired while compressing the significant areas. Otherwise, there may be difficulties or misdiagnosis in the treatment planning. With the lossless image compression technique, it is possible to preserve the entire pixel data while reducing the image size. The disadvantage of this technique is that it does not gain high memory size due to the low compression performance. On the other hand, higher compression ratios can be obtained by compromising redundant data with the lossy compression technique. Loss of image quality seems to be a risky condition in terms of correct diagnosis, but it is possible to reach acceptable compression rates and control data loss by setting appropriate parameters without losing diagnostic information. Adaptive image compression (AIC) is a hybrid technique that combines both lossless and lossy image compression techniques [2,3]. When applying AIC, it is primarily necessary to detect regions of interest in order to determine which regions will be compressed as lossy or lossless. After determining the focused or noticeable regions of radiologist on graph, it is possible to sort and adaptively compress these regions by importance. If the first order region is considered as the area containing the most information for the diagnosis, lossless compression can be applied here so as to avoid loss of detail. If there are second, third, and continuing order regions of interest, it may be preferable to adaptively compress these areas with little loss. Non-ROI (Non-Region of Interest) parts can be considered as less important or healthy areas in the diagnostic sense. Therefore, higher compression ratios can be achieved by compromising more details for these regions. After applying AIC, reconstructed image should be evaluated as sufficient and acceptable by the physician in terms of diagnostic information.Thus, the compressed images are recommended to be evaluated with subjective criteria in addition to objective criteria.The most common used objective criteria parameters for evaluating the compression performance are Compression Ratio (CR), Bits per Pixel (BPP), Mean Square Error (MSE) and the Peak Signal to Noise Ratio (PSNR). Compression ratio is defined as the

Sleep apnea is a common respiratory disorder during sleep. It is characterized by shallow or no breathing during sleep for at least 10 seconds. Decrease in sleep quality may effect the next day daily routine unfavorably. In some cases apnea period (not breathing interval) can last more than 30 seconds causing fatal outcomes. 14% of men and 5% of women suffer from Obstructive Sleep Apnea (OSA) in United States. Patients may face apnea for more than 300 times in a single overnight sleep. Polysomnography (PSG) is a multi-parametric recording of biophysiological changes, having Snorring, SpO2, Nasal Airflow EEG, EMG, ECG signals, performed in sleep study laboratories.In this study, a fully automatic apnea detection algorithm is mentinoed and an early warning system is proposed to predict OSA episodes by extracting time-series features of pre-OSA periods and regular respiration using nasal airflow signal. Extracted features are then reduced by RANSAC and entropy based approaches to improve the performance of prediction algorithm. Support vector machines (SVM), one of the commonly used classification algorithms in medical applications, k-Nearest Neighbor and a modified Linear Regression are implemented for learning and classification of nasal airflow signal episodes. The results show that OSA episodes are predicted with 86.9% of accuracy and 91.5% of sensitivity, 30 seconds before patient faces apnea. By the use of predicting an apnea episode before happening, it is possible to prevent patient to face apnea by early warning which can minimize the possible health risks.