OBSTRUCTIVE SLEEP APNEA DIAGNOSIS WITH APNEA EVENT DETECTION IN SNORING SOUND USING a CONDITIONAL RANDOM FIELD

Obstructive Sleep Apnea Diagnosis With Apnea Event Detection in Snoring Sound Using a Conditional Random Field He Lian (A0068205J) B.Sc Of Computer Science Peking University 2010 A Thesis Submitted For The Degree of Master of Science Department of Computer Science School of Computing National University of Singapore 2012 Abstract Obstructive Sleep Apnea (OSA) has become increasingly prevalent throughout the world in recent decades, but its proper diagnosis is severely constrained by the limited accessibility of polysomnography (PSG) facilities To resolve this problem, researchers investigated the potential of OSA diagnosis by using snore-related sounds However, most existing approaches to OSA diagnosis analyze snore episodes or silence episodes individually In this thesis, we propose a method to identify apnea events by incorporating ISPJ and F1 lables and learning the relation among these sequential acoustic signal components using a conditional random field Compared with three existing methods, the proposed method exhibits the best performance by achieving a sensitivity of 92.31% and a specificity of 80% under the threshold of apnea index set to Moreover, the number of apnea events detected by our approach effectively approximates the actual one reported by PSG, which makes the proposed method a potential alternative for manual annotation Based on the proposed method, a prototype named Mobile Obstructive Sleep Apnea Diagnosis is implemented on a mobile device Validation results demonstrate the prototype’s effectiveness and efficiency The efficacy and portability of our system illustrate its promising potential for OSA screening in a home environment i Acknowledgment I would like to express my sincere gratitude to Dr Wang Ye for his guidance and encouragement, to Dr Sim Khe Chai for his generous support, and to Dr Khoo See Meng for his cooperation in the data collection I am also grateful to Lee Yue Ting and Liu Liu for their assistance Lastly, I would like to show my appreciation to Fang Haotian for proofreading my thesis, as well as to everyone who has helped me along the way Thank you ii Contents Abstract i Acknowledgment ii List of Figures vi List of Tables vii Introduction 1.1 Motivation 1.2 Contributions 1.3 Organization Literature Survey 2.1 OSA Diagnosis With Snore Sound Analysis 2.2 Conditional Random Field 11 2.3 Portable OSA Diagnosis System 12 Apnea Events Detection using CRF 15 3.1 Data Collection 15 3.2 Automatic Segmentation 17 3.3 Apnea Event Detection using CRF 18 3.3.1 CRF Briefing 18 3.3.2 Association from respiratory events to silence episodes 20 3.3.3 Clique for CRF 21 3.3.4 Observation Extraction for CRF Training and Testing 22 3.3.5 Observation conjunction 25 iii Experiments 27 4.1 Experimental Parameters 27 4.2 Training Process 28 4.2.1 Parameter Determination 28 4.2.2 Training for CRF Model 29 4.3 Testing with CRF 30 4.4 Performance of Respiratory Event Detection 31 4.5 Comparison With Existing Diagnostic Methods 37 4.5.1 Comparison With the Snore-Episode-Related Methods 38 4.5.2 Comparison With the Respiratory-Event-Related Method 40 Mobile Obstructive Sleep Apnea Diagnosis 44 5.1 Recording of Snore-related Signal 45 5.2 Optimization of Audio Processing 46 5.2.1 Reduce Time Complexity of Audio Processing 47 5.2.2 Avoid Redundant Audio Analysis 49 5.3 Estimation of Total Sleeping Time 50 5.4 Real-time Diagnosis 50 5.5 Offline Diagnosis 51 5.6 Validation Experiments 52 5.6.1 Specification of Mobile Device 52 5.6.2 Performance Validation on Mobile Device 53 5.6.3 Efficiency Experiments 54 Conclusion and Future Work 56 Bibliography 58 iv A Appendix 66 v List of Figures 3.1 Flowchart of OSA diagnosis system with CRF 16 3.2 Association from respiratory annotations to silence episodes 20 3.3 Clique for CRF 22 4.1 ROC analysis for the threshold of F1 28 4.2 Effect of CRF on apnea event detection 31 4.3 Comparison between AENPSG and AENCRF 33 4.4 Comparison between AICRF and AIPSG 35 4.5 ROC curve for the percentage cutoff PPth of snores labeled with ISPJ 38 4.6 ROC curve for the percentage cutoff FPth of snores labeled with abnormal F1 38 4.7 Covariance between AHIPSG and PISPJ , PF1 and AICRF 40 4.8 Comparison between REN detected by EPD and by PSG 41 4.9 Covariance between AHIEPD and AHIPSG and that between AHIPSG and AICRF 42 5.1 MOSAD prototype 44 5.2 Recording audio queue 46 5.3 Cepstrum calculation 47 5.4 Extraction of pitch and formants from cepstrum 49 5.5 Playback audio queue 52 vi List of Tables 2.1 Comparison among three levels of portable OSA monitoring 13 2.2 Comparison of five portable OSA diagnosis systems 14 3.1 Duration label for silence and snore episodes 23 3.2 observation conjunction used in CRF model 25 4.1 Parameter setting for experiments 27 4.2 Statistics of subjects used for CRF training 29 4.3 Statistics of subjects used for CRF testing 30 4.4 Comparison of AENPSG and AENCRF and that of HENPSG and HENCRF 32 4.5 Performance of OSA diagnosis using CRF 34 4.6 Information of incorrectly categorized subjects 36 4.7 Comparison between AENPSG and correct AENCRF 36 4.8 Performance of OSA diagnosis using ISPJ, F1, and CRF 39 4.9 Comparison between OSA diagnosis results of EPD and CRF 42 5.1 Statistics to estimate total sleeping time 50 5.2 Specifications of the iPod Touch 53 5.3 Performance of OSA diagnosis using CRF on iOS 53 A.1 Information about subjects used in experiments 66 vii Introduction 1.1 Motivation Obstructive Sleep Apnea (OSA) is the most common sleep-related breathing disorder It is characterized by the total or partial obstruction of the upper airway during sleep, accompanied by repetitive cessation of respiratory airflow and frequent premature arousals Untreated OSA reduces the quality of sleep and increases the risk of heart disease, cognitive impairment, high blood pressure and stroke The loss of restorative sleep causes sleepiness during the day and contributes to the rising number of motor accidents [18, 38] OSA has become increasingly prevalent throughout the world in recent decades In India, the country with the second biggest population, 7.5% of men suffer from OSA [49] In the United States, an estimated 9% of middle-aged women and 24% of middleaged men have at least mild OSA [54] In Singapore, about 15% of the population is also estimated to be at risk [38] With the spread of the obesity epidemic, the incidents of OSA will continue to rise Polysomnography (PSG), which monitors airflow, blood oxygen saturation, brain activity (EEG), heart rhythm (ECG), eye movements (EOG), and muscle activity (EMG), is the standard diagnostic test for OSA However, it is complicated, expensive, and laborintensive Every PSG test attaches almost 20 sensors to the subject to monitor numerous body functions, and it costs around S$1000 and also requires professional technicians to stay an entire night to complete the diagnosis In addition, the scarcity of PSG facilities results in severely limited accessibility and considerable waiting time In Singapore, there are only two available sleeping laboratories, and the waiting time for a PSG is around three months These limitations may result in the under-diagnosis and under- treatment of millions of potential OSA patients In fact, it is estimated that more than 80% of affected individuals remains undiagnosed [19] Given the increasing prevalence of OSA and the limitations of PSG, researchers have investigated alternative diagnostic tools In medical field, questionnaires, such as the most famous Berlin Questionnaire [31, 11], the four-question STOP Questionnaire [10] and the clinical prediction model which are specifically derived for Singapore population in [27], were validated to be capable of predicting OSA These questionnaires collect diagnostic information including age, gender, the occurrence and frequnency of waking up in the night, sleepiness in the daytime etc., and then the related information is analyzed to product a probability of being OSA patients However, the answers of most questionnaires are objective and may require the assessment from patients’ accompany These two factors seriously affect the accuracy and feasibility of these medical prediction methods In computer science field, researchers explored the potential of OSA diagnosis using other modalities, such as nasal airway pressure [41], blood oxygen saturation [4], heart rate [40], and snore sound [3, 7, 16, 21, 26, 32, 33, 48, 52] The first three modalities, though highly correlated with OSA diagnosis, require specific sensors to finish the collection Specifically, nasal airway pressure needs to be measured by a sensor on the philtrum; blood oxygen saturation is usually measured by the pulse oximetry which should be placed on a thin part of patient’s body; the measurement of heart rate requires the famous medical technique named electrocardiograph (ECG) which usually connects several sensors to the patient’s body These constrains obstruct them to be widely applied in OSA home screening Snoring, the earliest manifestation of upper airway abnormalities, is strongly associated with OSA, affecting 70% to 95% of OSA patients [42] The snoring sound is generated by the vibration of soft tissues or the collapse of the upper airway due to air2 read into buffer Audio Queue Service Disk ! k buffer queue loudspeaker (a) Fill buffer by reading data from disk read into buffer Disk callback func Audio Queue Service ! k buffer queue loudspeaker (b) Playback via callback function Figure 5.5: Playback audio queue The ISPJ method calculates the percentage of snore episodes labeled with ISPJ Although there are two CRF methods, offline diagnosis accesses snoring sound in a different manner It adopts the Extended Audio File Service to sequentially read a chunk, which is much larger than the size of an Audio Queue buffer, from the recorded audio each time Each chunk is first segmented into non-snore, snore, and silence episodes, from which diagnostic information are then extracted When the entire piece of data is processed, a merge step is taken to link the boundaries between chunks, and the diagnostic information is then fed to the CRF to identify the apnea events 5.6 5.6.1 Validation Experiments Specification of Mobile Device The validation experiment uses an iPod Touch 4G with the following specifications: 52 installed OS CPU Storage Memory iPod Touch 4g iOS 5.0 ARM Cortex-A8 Apple A4 GHz (underclocked to 800 MHz) 32 GB flash memory 256 MB DRAM Table 5.2: Specifications of the iPod Touch As shown in Table 5.2, the iPod Touch’s memory capacity and CPU processing speed are the major limitations for OSA diagnosis and should thus be considered when designing mobile applications 5.6.2 Performance Validation on Mobile Device To evaluate the performance of the MOSAD prototype on mobile devices, all of the 18 pieces of recording are tested on the iPod Touch using real-time CRF diagnosis and offline CRF diagnosis, both of which perform consistently in classification Table 5.3 shows the performance of CRF implemented on PC (CRFPC ) and the one implemented on iOS (CRFiOS ) Optimal threshold TP FP FN TN Sensitivity Specificity CRFiOS (real-time diagnosis and offline diagnosis) AI = CRFPC 69.23% 100% 12 1 92.31% 80% AI = Table 5.3: Performance of OSA diagnosis using CRF on iOS Compared to CRFPC , CRFiOS performs better in identifying healthy subjects, but is less effective in identifying OSA patients Given the limitations of the iPod Touch’s 53 computing power, this discrepancy may arise from the following adaptations: • Different thresholds for segmentation In CRFiOS , the signal is recorded in linear PCM, which represents each sample with a 16-bit integer In contrast, the signal in CRFPC is transformed into a wave file, and each sample is a float number between and -1 The inconsistent representation of samples causes the derivation of different thresholds, consequently affecting the performance of segmentation • Different chunk sizes incoming signals are read and processed in 25M (fiveminute) chunks in CRFPC but 10M chunks in CRFiOS due to the limited memory capacity of mobile devices The chunk size will affect the segmentation because small chunks introduce more fragments • Different pitch extraction and F1 determination methods CRFPC utilizes the YIN algorithm and the LPC-based method to determine pitch and formant, respectively, but CRFiOS replaces these two methods with cepstrum calculation for efficiency These two methods may yield different pitch and formant values, which affects the training of CRF model and the inference of CRF in apnea event detection • Different TST CRFPC adopts the actual TSTPSG while CRFiOS estimates TST based on the duration of recording and the statistics collected from existing recordings The gap between the TSTPSG and the eTST affects the calculation of AHI 5.6.3 Efficiency Experiments In addition to performance, efficiency is a major factor in the mobile application design The most time-consuming part of the prototype is the calculation of pitch and 54 formant Using the original pitch determination and formant extraction method, offline diagnosis on one piece of recording can be completed in approximately one hour on the simulator on PC, but when running on an actual mobile device, it costs considerably more time As discussed before, even a three-minute clip requires almost 20 minutes to process This renders offline diagnosis using complete overnight recordings unfeasible However, after the optimization of audio processing, the average time for offline CRF diagnosis is 88±44 minutes, which is completely feasible and far more reasonable considering the long duration of a whole-night recording For real-time CRF diagnosis, the bottleneck is not time but power usage A fully charged iPod Touch can power three to four overnight recordings, but does not last even one night with real-time diagnosis Thus, the device must stay plugged in during the process The large amount of calculation for pitch and formant accounts for such power consumption Methods with less calculation should be investigated in the future 55 Conclusion and Future Work In this thesis, we proposed a relational learning method to detect apnea events using CRF, a well-known sequential labeling technique The proposed method combines features from the time domain, which capture the typical event pattern for apnea, and features from the frequency domain, which carry useful information about snore source and upper airway abnormalities To identify respiratory events, CRF determines the relations of these features among adjacent acoustic signal components, including silence episodes and snore episodes Experiments demonstrated the effectiveness of our method in apnea event detection Furthermore, compared with two existing snore-episode-related and one respiratory-events-related diagnostic methods, CRF performed better in the classification of OSA patients and healthy subjects Our proposed approach thus has the potential to relieve sleep technicians of the burden of manual annotation We also developed MOSAD, a prototype of OSA diagnosis on iOS, which can record snoring sound, provide diagnoses in both real-time and offline mode, and enable users to view past records and diagnosis results Validation experiments showed that both real-time and offline diagnosis could effectively identify healthy subjects but had room for improvement when it came to identifying OSA patients Experiments also demonstrated the reasonable efficiency of offline diagnosis Due to limited time and human resources, validation experiments were not conducted in the home environment However, the performance based on the recording collected from the sleeping laboratory shows MOSAD’s promise as an effective home screening tool for OSA Although the proposed method has been shown to effectively diagnose OSA, and preliminary validation demonstrates the potential of MOSAD prototype as a OSA home screening tool, we can still make our method and our prototype more effective and applicable by: 56 • Improving the accuracy of apnea detection Currently, CRF only uses four observations for the training and inference, and it only learns the relation between snore episodes and silences episodes Additional features and relation among other acoustic signal components, such as breath episodes, can be investigated • Investigating features to identify hypopnea events The CRF method cannot effectively detect hypopnea events because no obvious pattern for hypopnea has been observed in snoring signals Further research on finding typical patterns and features of hypopnea events in snoring sound should be carried out in the future • Validating the performance in home environment Although the MOSAD prototype is validated with data recorded in the ideal environment of sleeping laboratory, home environment may contain various background noise The squeaking of the bed, noise from the air conditioners, the clocks, rotating fans and sounds produced by the bed partner may interfere with the OSA diagnosis • Applying noise reduction techniques The data we used was recorded in the sleeping laboratory, which was with noise level intentionally reduced In actual home environments, noise reduction techniques such as time-frequency filter and source separation should be applied to pre-process the signal before further analyses 57 References [1] U R Abeyratne, C K K Patabandi, and K Puvanendran Pitch-jitter analysis of snoring sounds for the diagnosis of sleep apnea 2:2072–2075 vol.2, 2001 [2] Udantha Abeyratne, Asela Karunajeewa, and Craig Hukins Mixed-phase modeling in snore sound analysis Medical and Biological Engineering and Computing, 45(8):791–806, August 2007 [3] Udantha R Abeyratne, Ajith S Wakwella, and Craig Hukins Pitch jump probability measures for the analysis of snoring sounds in apnea Physiological Measurement, 26(5):779–798, October 2005 [4] D Alvarez, R Hornero, M García, F del Campo, and C Zamarrón Improving diagnostic ability of blood oxygen saturation from overnight pulse oximetry in obstructive sleep apnea detection by means of central tendency measure Artificial intelligence in medicine, 41(1):13, 2007 [5] I Ayappa and D.M Rapoport The upper airway in sleep: physiology of the pharynx Sleep Medicine Reviews, 7(1):9–33, 2003 [6] R Beck, M Odeh, A Oliven, and N Gavriely The acoustic properties of snores European Respiratory Journal, 8(12):2120–2128, 1995 [7] Cavusoglu, M., Ciloglu, T., Serinagaoglu, Y., Kamasak, M., Erogul, O., Akcam, and T Investigation of sequential properties of snoring episodes for obstructive sleep apnoea identification Physiological Measurement, 29(8):879–898, August 2008 58 [8] Cavusoglu, M., Kamasak, M., Erogul, O., Ciloglu, T., Serinagaoglu, Y., Akcam, and T An efficient method for snore/nonsnore classification of sleep sounds Physiological Measurement, 28(8):841–853, August 2007 [9] D.G Childers, D.P Skinner, and R.C Kemerait The cepstrum: A guide to processing Proceedings of the IEEE, 65(10):1428–1443, 1977 [10] F Chung, B Yegneswaran, P Liao, S.A Chung, S Vairavanathan, S Islam, A Khajehdehi, and C.M Shapiro Stop questionnaire: a tool to screen patients for obstructive sleep apnea Anesthesiology, 108(5):812–821, 2008 [11] F Chung, B Yegneswaran, P Liao, S.A Chung, S Vairavanathan, S Islam, A Khajehdehi, and C.M Shapiro Validation of the berlin questionnaire and american society of anesthesiologists checklist as screening tools for obstructive sleep apnea in surgical patients Anesthesiology, 108(5):822–830, 2008 [12] A De Cheveigné and H Kawahara Yin, a fundamental frequency estimator for speech and music J Acoust Soc Am, 111(4):1917–1930, 2002 [13] D DeCaprio, J.P Vinson, M.D Pearson, P Montgomery, M Doherty, and J.E Galagan Conrad: gene prediction using conditional random fields Genome research, 17(9):1389–1398, 2007 [14] Charalampos Doukas, Theodoros Petsatodis, and Ilias Maglogiannis MORFEAS: A Non-Invasive System for automated Sleep Apnea Detection utilizing Snore Sound Analysis In Ratko Magjarevic, Panagiotis D Bamidis, and Nicolas Pallikarakis, editors, XII Mediterranean Conference on Medical and Biological Engineering and Computing 2010, volume 29 of IFMBE Proceedings, chapter 30, pages 119–123 Springer Berlin Heidelberg, Berlin, Heidelberg, 2010 59 [15] Duckitt, W.D., Tuomi, S.K., Niesler, and T.R Automatic detection, segmentation and assessment of snoring from ambient acoustic data Physiological Measurement, 27(10):1047–1056, October 2006 [16] Takahiro Emoto, Udantha R Abeyratne, Masatake Akutagawa, Asela S Karunajeewa, Shinsuke Konaka, and Yohsuke Kinouchi Comparing benign snores with apneic snores based on high frequency analysis In Ratko Magjarevic, Olaf Dössel, and Wolfgang C Schlegel, editors, World Congress on Medical Physics and Biomedical Engineering, September - 12, 2009, Munich, Germany, volume 25/7 of IFMBE Proceedings, chapter 227, pages 820–823 Springer Berlin Heidelberg, Berlin, Heidelberg, 2009 [17] R Ferber, R Millman, M Coppola, J Fleetham, C.F Murray, C Iber, V McCall, G Nino-Murcia, M Pressman, M Sanders, et al Portable recording in the assessment of obstructive sleep apnea asda standards of practice Sleep, 17(4):378, 1994 [18] W.W Flemons, M.R Littner, J.A Rowley, P Gay, W.M.D Anderson, D.W Hudgel, R.D McEvoy, and D.I Loube Home diagnosis of sleep apnea: A systematic review of the literature* Chest, 124(4):1543–1579, 2003 [19] GJ Gibson Obstructive sleep apnoea syndrome: underestimated and undertreated British medical bulletin, 72(1):49–64, 2004 [20] X He, R.S Zemel, and M.A Carreira-Perpinán Multiscale conditional random fields for image labeling In Computer Vision and Pattern Recognition, 2004 CVPR 2004 Proceedings of the 2004 IEEE Computer Society Conference on, volume 2, pages II–695 IEEE, 2004 60 [21] Michael Herzog, Andreas Schmidt, Thomas Bremert, Beatrice Herzog, Werner Hosemann, and Holger Kaftan Analysed snoring sounds correlate to obstructive sleep disordered breathing European archives of oto-rhino-laryngology, 265(1):105–113, January 2008 [22] L Hou, S Xie, S Kai, and W Song Detection of osahs using only time-domain property of snoring signal In Multimedia Technology (ICMT), 2011 International Conference on, pages 519–522 IEEE, 2011 [23] R Jane, J A Fiza, J Sola-Soler, S Blanch, P Artis, and J Morera Automatic snoring signal analysis in sleep studies pages 366–369, 2003 [24] M.A Kammoun, D Gargouri, M Frikha, and A Ben Hamida Cepstral method evaluation in speech formant frequencies estimation In Industrial Technology, 2004 IEEE ICIT’04 2004 IEEE International Conference on, volume 3, pages 1612–1616 IEEE, 2004 [25] A S Karunajeewa, U R Abeyratne, and C Hukins Pitch/Airway-Response and Cepstral Analysis of Snore Sounds for the Non-Contact Screening of Sleep Apnea 25/4:2295–2298, 2010 [26] A S Karunajeewa, U R Abeyratne, and C Hukins Multi-feature snore sound analysis in obstructive sleep apnea–hypopnea syndrome Physiological measurement, 32:83, 2011 [27] S.M Khoo, H.K Poh, Y.H Chan, W.J Ngerng, D.X Shi, and TK Lim Diagnostic characteristics of clinical prediction models for obstructive sleep apnea in different clinic populations Sleep and Breathing, 15(3):431–437, 2011 [28] R Klinger and K Tomanek Classical probabilistic models and conditional random fields TU, Algorithm Engineering, 2007 61 [29] J Lafferty, A McCallum, and F.C.N Pereira Conditional random fields: Probabilistic models for segmenting and labeling sequence data 2001 [30] Z Lan, A Itoi, M Takashima, M Oda, and K Tomoda Difference of pharyngeal morphology and mechanical property between osahs patients and normal subjects Auris Nasus Larynx, 33(4):433–439, 2006 [31] N.C Netzer, R.A Stoohs, C.M Netzer, K Clark, K.P Strohl, et al Using the berlin questionnaire to identify patients at risk for the sleep apnea syndrome Annals of internal medicine, 131(7):485, 1999 [32] A K Ng, T S Koh, U R Abeyratne, and K Puvanendran Investigation of obstructive sleep apnea using nonlinear mode interactions in nonstationary snore signals Annals of biomedical engineering, 37(9):1796–1806, 2009 [33] Andrew K Ng, T S Koh, Eugene Baey, and K Puvanendran Diagnosis of Obstructive Sleep Apnea using Formant Features of Snore Signals In Ratko Magjarevic, R Magjarevic, and J H Nagel, editors, World Congress on Medical Physics and Biomedical Engineering 2006, volume 14 of IFMBE Proceedings, chapter 230, pages 967–970 Springer Berlin Heidelberg, Berlin, Heidelberg, 2007 [34] Andrew K Ng, Tong S Koh, Eugene Baey, Teck H Lee, Udantha R Abeyratne, and Kathiravelu Puvanendran Could formant frequencies of snore signals be an alternative means for the diagnosis of obstructive sleep apnea? Sleep Medicine, 9(8):894–898, December 2008 [35] Ashida Nobuyuki, Nasu Yasuhiro, Teshima Taiki, Yamakawa Miyae, Makimoto Kiyoko, and Higashi Terumasa Trial of measurement of sleep apnea syndrome with sound monitoring and SpO2 at home pages 66–69, December 2009 62 [36] A.M Noll Cepstrum pitch determination The journal of the acoustical society of America, 41:293, 1967 [37] Carole Planès, Michel Leroy, Bouach, El, Franck Digne, Florence de Roquefeuil, and Bernadette Raffestin Home diagnosis of obstructive sleep apnoea in coronary patients: validity of a simplified device automated analysis Sleep and Breathing [38] K Puvanendran and K L Goh From snoring to sleep apnea in a Singapore population Sleep research online : SRO, 2(1):11–14, 1999 [39] L.R Rabiner A tutorial on hidden markov models and selected applications in speech recognition Proceedings of the IEEE, 77(2):257–286, 1989 [40] F Roche, J.M Gaspoz, P Minini, V Pichot, D Duverney, F Costes, J.R Lacour, J.C Barthélémy, et al Screening of obstructive sleep apnea syndrome by heart rate variability analysis Circulation, 100(13):1411–1415, 1999 [41] J.I Salisbury and Y Sun Rapid screening test for sleep apnea using a nonlinear and nonstationary signal processing technique Medical engineering & physics, 29(3):336–343, 2007 [42] D Schlosshan and M Elliott Sleep* 3: Clinical presentation and diagnosis of the obstructive sleep apnoea hypopnoea syndrome Thorax, 59(4):347, 2004 [43] J Sola-Soler, R Jane, J A Fiz, and J Morera Towards automatic pitch detection in snoring signals 4:2974–2976 vol.4, 2000 [44] J Sola-Soler, R Jane, J A Fiz, and J Morera Snoring sound intensity study with ambient and tracheal microphones 2:2032–2035 vol.2, 2001 63 [45] J Sola-Soler, R Jane, J A Fiz, and J Morera Pitch analysis in snoring signals from simple snorers and patients with obstructive sleep apnea 2:1527–1528 vol.2, 2002 [46] J Sola-Soler, R Jane, J A Fiz, and J Morera Spectral envelope analysis in snoring signals from simple snorers and patients with Obstructive Sleep Apnea pages 2527–2530, 2003 [47] J Sola-Soler, R Jane, J A Fiz, and J Morera Variability of snore parameters in time and frequency domains in snoring subjects with and without Obstructive Sleep Apnea pages 2583–2586, 2005 [48] Jordi Sola-Soler, Raimon Jane, Jose A Fiz, and Jose Morera Automatic classification of subjects with and without Sleep Apnea through snoring analysis pages 6093–6096, August 2007 [49] Z.F Udwadia, A.V Doshi, S.G Lonkar, and C.I Singh Prevalence of sleepdisordered breathing and sleep apnea in middle-aged urban indian men American journal of respiratory and critical care medicine, 169(2):168, 2004 [50] S.B Wang, A Quattoni, L.P Morency, D Demirdjian, and T Darrell Hidden conditional random fields for gesture recognition In Computer Vision and Pattern Recognition, 2006 IEEE Computer Society Conference on, volume 2, pages 1521– 1527 Ieee, 2006 [51] P.R Westbrook, D.J Levendowski, M Cvetinovic, T Zavora, V Velimirovic, D Henninger, and D Nicholson Description and validation of the apnea risk evaluation system* Chest, 128(4):2166–2175, 2005 [52] A Yadollahi and Z Moussavi Acoustic obstructive sleep apnea detection pages 7110–7113, September 2009 64 [53] M Yin, S Miyazaki, Y Itasaka, Y Shibata, T Abe, A Miyoshi, K Ishikawa, and K Togawa A preliminary study on application of portable monitoring for diagnosis of obstructive sleep apnea Auris Nasus Larynx, 32(2):151–156, 2005 [54] T Young, M Palta, J Dempsey, J Skatrud, S Weber, and S Badr The occurrence of sleep-disordered breathing among middle-aged adults The New England journal of medicine, 328(17):1230–1235, April 1993 65 A Appendix Group training testing Subject AHI AEN HEN AI HI S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 S27 S28 33.6 75 69.6 94.7 40.8 18.4 14.5 55.9 1.9 24.8 86.1 90.4 43.3 68.8 98.6 46.7 0.5 22 12.6 27.8 40.2 72 15.7 57.3 6.1 10.3 15.1 2.5 100 356 263 44 175 16 44 88 392 274 138 359 307 137 11 18 25 51 355 36 165 45 65 107 49 153 392 107 62 61 272 11 47 23 202 89 16 320 162 114 46 71 174 33 57 189 30 28 16.2 65.9 43.9 9.4 25.3 0.9 8.1 0.2 16.2 81.5 52.4 26.6 65.7 48.6 21.6 1.9 3.5 7.4 8.7 65.8 6.1 26.7 0.7 8.7 10.5 1.1 17.4 9.1 25.7 85.3 15.5 17.6 11.5 47.8 1.7 8.6 4.6 38 16.7 2.9 50 25.1 0.5 20 20.4 28.8 6.1 9.6 30.6 5.4 1.6 4.5 1.4 TST (mins) 369.5 324 355 275 413 211.5 298.5 326.5 386.5 314.5 288.5 314 304.5 327 379 375.5 353.5 338.5 305 203 352 323.5 355 371 333.5 309.5 373 377.5 longest AE (mins) 61.7 49.8 78.6 48.1 60.6 22.5 102.4 35.4 16.9 83.6 69.7 52.6 43.6 63.9 85.6 73.4 23.9 43 39.8 49.6 96.8 30.2 59 43.3 68.8 46.4 51.3 Table A.1: Information about subjects used in experiments 66 longest HE (mins) 64 40.3 96.8 85.3 90.4 60.3 81.1 48.7 41.3 72.2 60 56.3 45.6 40.3 39.7 48 41.1 80.3 66.1 60.3 90.8 67.3 42.6 76.2 51.2 47.5 37.4 79.7 [...]... during routine PSGs in the sleeping laboratory of the National University Hospital in Singapore Corresponding 15 Annotation PSG Apnea and Hypopnea annotation Snore Signal silence episodes non-snore episodes snore episodes events label segmentation class label duration label ISPJ label F1 label train data observation for CRF CRF model test data apnea events apnea events detection Figure 3.1: Flowchart... episodes Apnea and hypopnea events annotated by technicians are then associated with detected silence episodes to generate an event label • CRF observation extraction In this part, observations for CRF are extracted from the components above, which, together with their event labels, serve as the training data for CRF • Apnea events detection A CRF model is trained to detect apnea events for OSA diagnosis. .. episodes make sequential labeling of respiratory events a valid and potentially highly effective approach to OSA diagnosis However, a manually annotated respiratory event does not strictly associate with an 11 individual snore episode or silence episode In fact, one manually annotated respiratory event may include several snore, breath, and silence episodes, and its boundaries are inconsistent with those... specifically apnea and hypopnea events In [22], Hou and Xie et al defined a respiratory event as an interval longer than 10 seconds between two adjacent snore events They attempted to detect respiratory events using a dynamic threshold for endpoint detection (EPD) of snore episodes Although their target parameter was directly related to the calculation of AHI, their definition and method had several weaknesses... from OSA patients Pitch-jitter analysis could classify snore episodes into AS class with 92.31% accuracy and BS class with 90.7% accuracy, suggesting that pitch might be a suitable candidate to identify apnea snores Abeyratne and Karunajeewa et al also designed an algorithm to segment snore-related -sound (SRS) into classes of pure breathing, silence, and voiced/unvoiced snores using pitch [3] SRS was first... episodes apnea silences or hypopnea silences According to the definition of apnea events, it is also the silence episodes longer than 10 seconds that most strongly associate with apnea events Therefore, in this paper, we associate a real apnea and hypopnea event to the longest silence episode between its start and end positions as shown in Figure 3.2 The associated silence episodes are labeled as apnea or... segmentation of snoring sound, feature selection and model selection Abeyratne and Karunajeewa et al contributed extensively to OSA diagnosis based on snore-related sound analyses In their early research, they carried out pitch-jitter analysis to separate the signal into benign snore (BS), apnea snore (AS), and speech [1] Benign snore was defined as a snore episode from healthy subjects while apnea snore... Hypopnea Index (AHI), which is defined as the number of respiratory events per sleeping hour Respiratory events consist of two types: apnea event and hypopnea event Apnea event refers to the complete cessation of nasal or oral airflow lasting for at least 10 seconds Hypopnea event refers to a segment with 50% reduction of airflow for at least 10 seconds and is accompanied by decreased blood oxygen saturation... snoring level, arterial oxygen saturation, pulse rate, head movement a brain monitor affixed to forehead III snoring sound, oronasal airflow, arterial oxygen saturation, pulse rate, body position, etc Stardust II device IV snore sound Hardware III tracheal breath, nasal flow, body position, arterial oxygen saturation, heart rate, etc recorder, analyzer Ashida’s System IV snore sound, SpO2 Questionnaire... be an OSA patient based on their defined measurement, but not tell him or her how sever the sleeping apnea is Meanwhile, the useful information contained in silence episodes does not catch much attention in the research of OSA diagnosis using snoring sound However, it is these silence episodes that are closely related to the clinical measurement of OSA OSA severity is clinically measured by the Apnea ... that validates the performance on apnea event detection Moreover, we have shown that automatic annotation of apnea events using CRF has the potential to replace manual annotation and reduce labor... on apnea event detection 31 120 Figure 4.2 shows the effect of apnea event detection using CRF on a two-minute snoring sound clip As shown in this figure, manually annotated apnea events are always... events label segmentation class label duration label ISPJ label F1 label train data observation for CRF CRF model test data apnea events apnea events detection Figure 3.1: Flowchart of OSA diagnosis