Cardiopal: Cardiac passive acoustic localization and mapping using2-D recording of heart sounda
Başlık çevirisi mevcut değil.
- Tez No: 65140
- Danışmanlar: DOÇ. DR. HALİ ÖZCAN GÜLÇÜR
- Tez Türü: Doktora
- Konular: Biyomühendislik, Bioengineering
- Anahtar Kelimeler: Heart Sounds, Digital Phonocardiography, Adaptive Filtering, Autoregressive Modeling, Pattern Recognition, Artificial Neural Network, Prony's Method, Microphone Array, Subspace-based Array Processing, Heart Sounds, Digital Phonocardiography, Adaptive Filtering, Autoregressive Modeling, Pattern Recognition, Artificial Neural Network, Prony's Method, Microphone Array, Subspace-based Array Processing
- Yıl: 1997
- Dil: İngilizce
- Üniversite: Boğaziçi Üniversitesi
- Enstitü: Fen Bilimleri Enstitüsü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 174
Özet
IV CARDIOPAL: CARDIAC PASSIVE ACOUSTIC LOCALIZATION AND MAPPING USING 2-D RECORDINGS OF HEART SOUNDS ABSTRACT In this thesis, a novel non-invasive system is developed as an adjunct diagnostic tool for cardiologists which is based on processing of diastolic heart sounds acquired using a 2-D passive acoustic array. The system consists of a specially designed 2-D passive acoustic array, instrumentation hardware and a 2-D array processing software. The array developed is a 105 mm x 105 mm x 40 mm planar array made using 16 miniature electret microphones mounted on a 4x4 grid. Data acquisition hardware of the system includes microphone amplifiers, 80-1000 Hz bandpass filters, ADCs and hardware interface to a PC. Basic assumptions are made about the near-field low frequency sources and a signal model is developed in accordance with these assumptions. It is employed in the Multiple Signal Characterization (MUSIC) method used for localization of the point sources. Locations of the sources were estimated by applying 2-D searches on a plane. Effectiveness of the method was demonstrated using extensive recordings at different SNR levels on phantoms constructed using small sound sources and a gel composition which mimic soft tissue. The system was also tested with success on adult subjects as well as on a six month old fetus. In the studies different“images”corresponding to the various distinct phases of the heartbeat; e.g., closure of the mitral and tricuspid valves, ejection of the blood (systole), closure of the aortic and pulmonary valves, early and late diastole etc., were obtained. It has also been shown that three dimensional“images”could be extracted using the same array data by evaluating a parameter related to the distance between the sources and the origin of the array. As the first phase of the study, a single channel prototype of the system has been tried in a real hospital setting. The system can be used in an ordinary hospital room since it is equipped with a very efficient adaptive noise cancellation scheme and does not require a special sound-proof room. Using the prototype system heart sounds from 84 subjects in a relatively low noise hospital room were recorded. During the recordings, ambient sound in the room and the subject's ECG were also acquired. Signals from the second sound channel were utilized to eliminate the ambient sounds from the isolated diastolic heart sounds using a frequency domain adaptive filter. Autoregressive (AR) model parameters were then estimated for each patient. Any difference in the AR parameters from normal and abnormal subjects, which recently had angiography, was investigated. Probability Density Functions (PDF) of the AR parameters were found to have significant differences between the two groups. AR parameters of the patients were then used to train a two layer perceptron. The two layer perceptron trained using the data obtained, correctly classified 67 percent of the cases not included in the training phase. A K-means classifier and a Bayes classifier were also implemented with the first having slightly better performance.As an additional by-product of the prototype system developed, a non-invasive method for systolic pressure monitoring was considered, based on the correlation between the systolic blood pressure measured at the aortic root and the average parameters of the damping sinusoidal model decomposition of the second heart sounds. This study was motivated by the knowledge that hypertensive patients have accentuated S2 sounds due to higher pressures in the aortic root. The closure sounds of the heart are transient in time and they damp in relatively short duration in a cardiac cycle. Therefore, they can be reconstructed well with a certain number of exponentially damping sinusoids using the Prony 's method. A linear correlation between the systolic blood pressure and the average parameters, the amplitude, damping parameter, the phase and the frequencies of the damped sinusoidal model were tested by calculating the Pearson coefficient. The damping parameter, the amplitude and the phase parameters have been found to have considerably high correlation. However, the frequencies of the sinusoids do not seem to be correlated with the blood pressure. The results are encouraging and show that the proposed approach may be accurate enough for long-term noninvasive systolic blood pressure monitoring in the aortic root, with much less discomfort to the patients than other methods.
Özet (Çeviri)
IV CARDIOPAL: CARDIAC PASSIVE ACOUSTIC LOCALIZATION AND MAPPING USING 2-D RECORDINGS OF HEART SOUNDS ABSTRACT In this thesis, a novel non-invasive system is developed as an adjunct diagnostic tool for cardiologists which is based on processing of diastolic heart sounds acquired using a 2-D passive acoustic array. The system consists of a specially designed 2-D passive acoustic array, instrumentation hardware and a 2-D array processing software. The array developed is a 105 mm x 105 mm x 40 mm planar array made using 16 miniature electret microphones mounted on a 4x4 grid. Data acquisition hardware of the system includes microphone amplifiers, 80-1000 Hz bandpass filters, ADCs and hardware interface to a PC. Basic assumptions are made about the near-field low frequency sources and a signal model is developed in accordance with these assumptions. It is employed in the Multiple Signal Characterization (MUSIC) method used for localization of the point sources. Locations of the sources were estimated by applying 2-D searches on a plane. Effectiveness of the method was demonstrated using extensive recordings at different SNR levels on phantoms constructed using small sound sources and a gel composition which mimic soft tissue. The system was also tested with success on adult subjects as well as on a six month old fetus. In the studies different“images”corresponding to the various distinct phases of the heartbeat; e.g., closure of the mitral and tricuspid valves, ejection of the blood (systole), closure of the aortic and pulmonary valves, early and late diastole etc., were obtained. It has also been shown that three dimensional“images”could be extracted using the same array data by evaluating a parameter related to the distance between the sources and the origin of the array. As the first phase of the study, a single channel prototype of the system has been tried in a real hospital setting. The system can be used in an ordinary hospital room since it is equipped with a very efficient adaptive noise cancellation scheme and does not require a special sound-proof room. Using the prototype system heart sounds from 84 subjects in a relatively low noise hospital room were recorded. During the recordings, ambient sound in the room and the subject's ECG were also acquired. Signals from the second sound channel were utilized to eliminate the ambient sounds from the isolated diastolic heart sounds using a frequency domain adaptive filter. Autoregressive (AR) model parameters were then estimated for each patient. Any difference in the AR parameters from normal and abnormal subjects, which recently had angiography, was investigated. Probability Density Functions (PDF) of the AR parameters were found to have significant differences between the two groups. AR parameters of the patients were then used to train a two layer perceptron. The two layer perceptron trained using the data obtained, correctly classified 67 percent of the cases not included in the training phase. A K-means classifier and a Bayes classifier were also implemented with the first having slightly better performance.As an additional by-product of the prototype system developed, a non-invasive method for systolic pressure monitoring was considered, based on the correlation between the systolic blood pressure measured at the aortic root and the average parameters of the damping sinusoidal model decomposition of the second heart sounds. This study was motivated by the knowledge that hypertensive patients have accentuated S2 sounds due to higher pressures in the aortic root. The closure sounds of the heart are transient in time and they damp in relatively short duration in a cardiac cycle. Therefore, they can be reconstructed well with a certain number of exponentially damping sinusoids using the Prony 's method. A linear correlation between the systolic blood pressure and the average parameters, the amplitude, damping parameter, the phase and the frequencies of the damped sinusoidal model were tested by calculating the Pearson coefficient. The damping parameter, the amplitude and the phase parameters have been found to have considerably high correlation. However, the frequencies of the sinusoids do not seem to be correlated with the blood pressure. The results are encouraging and show that the proposed approach may be accurate enough for long-term noninvasive systolic blood pressure monitoring in the aortic root, with much less discomfort to the patients than other methods.
Benzer Tezler
- A new hardware for cardiac passive acoustic localization
Kardiyak pasif akustik konumlandırma için yeni bir donanım tasarımı
YAHYA CİVELEK
Yüksek Lisans
İngilizce
2005
BiyomühendislikBoğaziçi ÜniversitesiBiyomühendislik Ana Bilim Dalı
DOÇ. DR. ÖZCAN GÜLÇÜR
- Açık kalp cerrahisinde kullanılan iki farklı kardiyopleji yönteminin koroner bypass operasyonunda böbrek fonksiyonları üzerine etkilerinin karşılaştırılması
Comparison of the effects on renal function in patients with coronary bypass surgery two different cardioplegia methods used in open heart surgery
ÖMER ARICI
Yüksek Lisans
Türkçe
2020
Göğüs Kalp ve Damar CerrahisiKahramanmaraş Sütçü İmam ÜniversitesiKalp ve Damar Cerrahisi Ana Bilim Dalı
DOÇ. DR. ERDİNÇ EROĞLU
- Tip 2 diyabetik hastalarda düzeltilmiş qt intervalinin kardiak sonlanımla ilişkisi
Relationship of corrected qt interval with cardiac outcomein type 2 diabetic patients
REMZİ YILDIZ
Tıpta Uzmanlık
Türkçe
2011
İç HastalıklarıSağlık BakanlığıDahiliye Ana Bilim Dalı
DR. ÖĞR. ÜYESİ MUSTAFA TEMİZEL
- Kalp yetersizliği hastalarının psikososyal uyumlarının değerlendirilmesi
The evaluation of psychosocial adaptation in heart failure patients
SEMİHA AKIN
Yüksek Lisans
Türkçe
2003
Hemşirelikİstanbul Üniversitesiİç Hastalıkları Hemşireliği Ana Bilim Dalı
PROF. DR. ZEHRA DURNA
- Kardiyak sendrom x'li hastalarda sol ventrikül global ve bölgesel sistolik fonksiyonlarının kolor kinesis dobutamin stres ekokardiyografi ile incelenmesi
Investigation of the left ventricular global and regional systolic functions in patients with cardiac syndrome x using color kinesis dobutamine stress echocardiography
ERKAN ÖZTEKİN
Tıpta Uzmanlık
Türkçe
2001
KardiyolojiTrakya ÜniversitesiKardiyoloji Ana Bilim Dalı
PROF. DR. GÜLTAÇ ÖZBAY