Using time-frequency features to recognize abnormal heart sounds

Hsuan Lin Her, Hung Wen Chiu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

Disease of the heart accounts for 6% of all death. Heart sound is a routine in physical examination clinically, and is sensitive in detecting a subset of heart diseases. In the current study, we build up a model to classify heart sounds for preclinical screening. Heart sounds are recorded under uncontrolled environment, and each sample can range from 5 to 120 seconds. This is a work raised by annual PhysioNet/CinC Challenge. Because of timing and tonic natures of heart events, we used time-frequency features to classify heart sounds in this study. Firstly, each heard sound recording was segmented into cycles using Springer's improved version of Schmidt's method. Each cardiac cycle was cut into 10 partitions and data points were obtained by zero-padding in each partition. Spectral features were extracted from each partition using fast-Fourier Transform (FFT) thus a 3,500 feature matrix was created. Using filter method, 40 features were selected for the final classifier. The average feature matrix of each cycle was then applied to a classification system using 2-means clustering and artificial neural network (ANN). By clustering the unsure class was recognized. The discrimination of normal and abnormal heart sound were performed by a well-trained ANN model. The results showed that our proposed method got a performance with an accuracy 86.5%, a sensitivity 84.4%, a specificity 86.9%. Here we show that classifying abnormal heart sound is a really difficult task due to the heterogeneity of 'abnormal events' and intra-sample deviation.

Original languageEnglish
Title of host publicationComputing in Cardiology Conference, CinC 2016
PublisherIEEE Computer Society
Pages1145-1147
Number of pages3
Volume43
ISBN (Electronic)9781509008964
Publication statusPublished - 2016
Event43rd Computing in Cardiology Conference, CinC 2016 - Vancouver, Canada
Duration: Sep 11 2016Sep 14 2016

Other

Other43rd Computing in Cardiology Conference, CinC 2016
CountryCanada
CityVancouver
Period9/11/169/14/16

Fingerprint

Heart Sounds
Acoustic waves
Cluster Analysis
Heart Diseases
Neural Networks (Computer)
Fourier Analysis
Sound recording
Physical Examination
Neural networks
Fast Fourier transforms
Screening
Classifiers

ASJC Scopus subject areas

  • Computer Science(all)
  • Cardiology and Cardiovascular Medicine

Cite this

Her, H. L., & Chiu, H. W. (2016). Using time-frequency features to recognize abnormal heart sounds. In Computing in Cardiology Conference, CinC 2016 (Vol. 43, pp. 1145-1147). [7868950] IEEE Computer Society.

Using time-frequency features to recognize abnormal heart sounds. / Her, Hsuan Lin; Chiu, Hung Wen.

Computing in Cardiology Conference, CinC 2016. Vol. 43 IEEE Computer Society, 2016. p. 1145-1147 7868950.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Her, HL & Chiu, HW 2016, Using time-frequency features to recognize abnormal heart sounds. in Computing in Cardiology Conference, CinC 2016. vol. 43, 7868950, IEEE Computer Society, pp. 1145-1147, 43rd Computing in Cardiology Conference, CinC 2016, Vancouver, Canada, 9/11/16.
Her HL, Chiu HW. Using time-frequency features to recognize abnormal heart sounds. In Computing in Cardiology Conference, CinC 2016. Vol. 43. IEEE Computer Society. 2016. p. 1145-1147. 7868950
Her, Hsuan Lin ; Chiu, Hung Wen. / Using time-frequency features to recognize abnormal heart sounds. Computing in Cardiology Conference, CinC 2016. Vol. 43 IEEE Computer Society, 2016. pp. 1145-1147
@inproceedings{d7b5ebf6c63f4933a224d3cf8f198428,
title = "Using time-frequency features to recognize abnormal heart sounds",
abstract = "Disease of the heart accounts for 6{\%} of all death. Heart sound is a routine in physical examination clinically, and is sensitive in detecting a subset of heart diseases. In the current study, we build up a model to classify heart sounds for preclinical screening. Heart sounds are recorded under uncontrolled environment, and each sample can range from 5 to 120 seconds. This is a work raised by annual PhysioNet/CinC Challenge. Because of timing and tonic natures of heart events, we used time-frequency features to classify heart sounds in this study. Firstly, each heard sound recording was segmented into cycles using Springer's improved version of Schmidt's method. Each cardiac cycle was cut into 10 partitions and data points were obtained by zero-padding in each partition. Spectral features were extracted from each partition using fast-Fourier Transform (FFT) thus a 3,500 feature matrix was created. Using filter method, 40 features were selected for the final classifier. The average feature matrix of each cycle was then applied to a classification system using 2-means clustering and artificial neural network (ANN). By clustering the unsure class was recognized. The discrimination of normal and abnormal heart sound were performed by a well-trained ANN model. The results showed that our proposed method got a performance with an accuracy 86.5{\%}, a sensitivity 84.4{\%}, a specificity 86.9{\%}. Here we show that classifying abnormal heart sound is a really difficult task due to the heterogeneity of 'abnormal events' and intra-sample deviation.",
author = "Her, {Hsuan Lin} and Chiu, {Hung Wen}",
year = "2016",
language = "English",
volume = "43",
pages = "1145--1147",
booktitle = "Computing in Cardiology Conference, CinC 2016",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Using time-frequency features to recognize abnormal heart sounds

AU - Her, Hsuan Lin

AU - Chiu, Hung Wen

PY - 2016

Y1 - 2016

N2 - Disease of the heart accounts for 6% of all death. Heart sound is a routine in physical examination clinically, and is sensitive in detecting a subset of heart diseases. In the current study, we build up a model to classify heart sounds for preclinical screening. Heart sounds are recorded under uncontrolled environment, and each sample can range from 5 to 120 seconds. This is a work raised by annual PhysioNet/CinC Challenge. Because of timing and tonic natures of heart events, we used time-frequency features to classify heart sounds in this study. Firstly, each heard sound recording was segmented into cycles using Springer's improved version of Schmidt's method. Each cardiac cycle was cut into 10 partitions and data points were obtained by zero-padding in each partition. Spectral features were extracted from each partition using fast-Fourier Transform (FFT) thus a 3,500 feature matrix was created. Using filter method, 40 features were selected for the final classifier. The average feature matrix of each cycle was then applied to a classification system using 2-means clustering and artificial neural network (ANN). By clustering the unsure class was recognized. The discrimination of normal and abnormal heart sound were performed by a well-trained ANN model. The results showed that our proposed method got a performance with an accuracy 86.5%, a sensitivity 84.4%, a specificity 86.9%. Here we show that classifying abnormal heart sound is a really difficult task due to the heterogeneity of 'abnormal events' and intra-sample deviation.

AB - Disease of the heart accounts for 6% of all death. Heart sound is a routine in physical examination clinically, and is sensitive in detecting a subset of heart diseases. In the current study, we build up a model to classify heart sounds for preclinical screening. Heart sounds are recorded under uncontrolled environment, and each sample can range from 5 to 120 seconds. This is a work raised by annual PhysioNet/CinC Challenge. Because of timing and tonic natures of heart events, we used time-frequency features to classify heart sounds in this study. Firstly, each heard sound recording was segmented into cycles using Springer's improved version of Schmidt's method. Each cardiac cycle was cut into 10 partitions and data points were obtained by zero-padding in each partition. Spectral features were extracted from each partition using fast-Fourier Transform (FFT) thus a 3,500 feature matrix was created. Using filter method, 40 features were selected for the final classifier. The average feature matrix of each cycle was then applied to a classification system using 2-means clustering and artificial neural network (ANN). By clustering the unsure class was recognized. The discrimination of normal and abnormal heart sound were performed by a well-trained ANN model. The results showed that our proposed method got a performance with an accuracy 86.5%, a sensitivity 84.4%, a specificity 86.9%. Here we show that classifying abnormal heart sound is a really difficult task due to the heterogeneity of 'abnormal events' and intra-sample deviation.

UR - http://www.scopus.com/inward/record.url?scp=85016104032&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85016104032&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85016104032

VL - 43

SP - 1145

EP - 1147

BT - Computing in Cardiology Conference, CinC 2016

PB - IEEE Computer Society

ER -