본 논문은 가중 퍼지소속함수 기반 신경망 (Neural Network with Weighted Fuzzy Membership functions, NEWFM)과 심박수 변이도(Heart Rate Variability, HRV)를 이용하여 우울증 진단알고리즘을 제안하고 있다. 본 알고리즘에서 사용할 NEWFM의 입력특징을 추출하기 위해서 주파수도메인 특징추출, 시간도메인 특징추출, 웨이블릿변환 특징추출, 포인케어변환 특징추출 방법을 이용하여 22개의 초기 HRV 특징들을 추출하였다. 또한 NEWFM에서 제공하는 비중복면적 분산측정법 (Non-overlap Area Distribution Measurement, NADM)에 의해 입력특징의 중요도를 평가하여 22개의 초기특징으로부터 중요도가 가장 높은 6개 최적입력특징을 선택하였다. 이 6개 특징을 이용하여 우울증을 진단한 결과는 95.8% 의 정확도를 나타내었다.
본 논문은 가중 퍼지소속함수 기반 신경망 (Neural Network with Weighted Fuzzy Membership functions, NEWFM)과 심박수 변이도(Heart Rate Variability, HRV)를 이용하여 우울증 진단알고리즘을 제안하고 있다. 본 알고리즘에서 사용할 NEWFM의 입력특징을 추출하기 위해서 주파수도메인 특징추출, 시간도메인 특징추출, 웨이블릿변환 특징추출, 포인케어변환 특징추출 방법을 이용하여 22개의 초기 HRV 특징들을 추출하였다. 또한 NEWFM에서 제공하는 비중복면적 분산측정법 (Non-overlap Area Distribution Measurement, NADM)에 의해 입력특징의 중요도를 평가하여 22개의 초기특징으로부터 중요도가 가장 높은 6개 최적입력특징을 선택하였다. 이 6개 특징을 이용하여 우울증을 진단한 결과는 95.8% 의 정확도를 나타내었다.
This paper presents an algorithm for depression diagnosis using the Neural Network with Weighted Fuzzy Membership functions (NEWFM) and heart rate variability (HRV). In the algorithm, 22 different features were initially extracted from the HRV signal by frequency domain, time domain, wavelet transfo...
This paper presents an algorithm for depression diagnosis using the Neural Network with Weighted Fuzzy Membership functions (NEWFM) and heart rate variability (HRV). In the algorithm, 22 different features were initially extracted from the HRV signal by frequency domain, time domain, wavelet transformed, and Poincar$\acute{e}$ transformed feature extraction methods; of these 6 optimal features were selected by significance evaluation using Non-overlap Area Distribution Measurement (NADM) based on NEWFM. The proposed algorithm uses these 6 optimal features to diagnose depression with an accuracy of 95.83%.
This paper presents an algorithm for depression diagnosis using the Neural Network with Weighted Fuzzy Membership functions (NEWFM) and heart rate variability (HRV). In the algorithm, 22 different features were initially extracted from the HRV signal by frequency domain, time domain, wavelet transformed, and Poincar$\acute{e}$ transformed feature extraction methods; of these 6 optimal features were selected by significance evaluation using Non-overlap Area Distribution Measurement (NADM) based on NEWFM. The proposed algorithm uses these 6 optimal features to diagnose depression with an accuracy of 95.83%.
* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.
제안 방법
In this study, initial twenty-two different features are extracted from the HRV signal by the frequency domain feature (FDF), time domain feature (TDF), wavelet transformed feature (WTF), and Poincaré transformed feature (PTF) extraction methods.
Based on the above, this paper proposed a depression diagnosis algorithm which only uses six features of HRV signal to diagnose depression based on neuro-fuzzy networks. In this study, initial twenty-two different features are extracted from the HRV signal by the frequency domain feature (FDF), time domain feature (TDF), wavelet transformed feature (WTF), and Poincaré transformed feature (PTF) extraction methods.
In this study, on the basis of a neuro-fuzzy network with a weighted fuzzy membership function (NEWFM)[8-12], a new depression diagnosis algorithm was used to distinguish depressed subjects from control subjects. NEWFM is a supervised classification neuro-fuzzy system, which can obtain the bounded sum of weighted fuzzy membership functions (BSWFMs)[8-12] of input features based on training processing.
The proposed depression diagnosis model includes 2 schemas, which are depression diagnosis and feature selection. In the feature selection schema, the initially 22 features included 4 FDFs, 7 TDFs, 9 WTFs, and 2 PTFs extracted from the HRV signal, from which the 6 optimal features were selected by NADM[8-12].
The proposed depression diagnosis model includes 2 schemas, which are depression diagnosis and feature selection. In the feature selection schema, the initially 22 features included 4 FDFs, 7 TDFs, 9 WTFs, and 2 PTFs extracted from the HRV signal, from which the 6 optimal features were selected by NADM[8-12]. The depression diagnosis schema has 4 steps, as shown in [Figure 2].
This study designed a experiment which used for HRV signals collection. In the experiment, HRV signals were collected from 24 subjects which wore a wireless Holter monitor.
Many research studies have examined the influence of emotions on the ANS utilizing the analysis of HRV[5-7]. This study designed a new MAC stimulus, which can evoke various emotions, such as happiness, joy, pain, stress, irritability, and fear. While each subject underwent the MAC test, he/she ate some soft marshmallows and super-sour candies, drank sweet juice, and blew up a balloon, among other tasks.
This study designed a new MAC stimulus, which can evoke various emotions, such as happiness, joy, pain, stress, irritability, and fear. While each subject underwent the MAC test, he/she ate some soft marshmallows and super-sour candies, drank sweet juice, and blew up a balloon, among other tasks. The MAC stimulus scenario is summarized in [Table 1].
In this study, 4 FDFs were selected according to the conclusions of other studies[15][16] and HRV measurement standards[17], and these were extracted via the corresponding extraction method. The FDFs can commonly be used for short-term recordings[17].
A total of 22 features were extracted using the FDF, TDF, WTF, and PTF extraction methods, as shown in [Table 2].
After the feature extraction process, the 22 features of the subjects were used as NEWFM’s input features to train NEWFM.
After the feature extraction process, the 22 features of the subjects were used as NEWFM’s input features to train NEWFM. During the training process, 22 BSWFMs of the input features were evaluated by NADM, which finds and counts the best input features among all the initial features using the evaluation methods [8-12]. The NEWFM was trained on the sampling data sets while the sampling data sets were tested (close test) 5,000 times; the result of the 5,000 tests appears in [Figure 4].
This paper uses 4 types of feature sets, which are 22 initial features and 6 optimal features for diagnosing depression. [Table 3] presents a comparison between the performance of the 22 initial features and the performance of the 6 optimal features on the 24 subjects' HRV signal sets.
Medical studies show a significant relationship between depression and HRV features[7]. This paper proposes a new depression diagnosis algorithm based on NEWFM. Six optimal features - VLF, SDNN, PNN100, Power_d3, Power_d2, and SD2 - were selected by NADM as input features of NEWFM to diagnose depression.
This paper proposes a new depression diagnosis algorithm based on NEWFM. Six optimal features - VLF, SDNN, PNN100, Power_d3, Power_d2, and SD2 - were selected by NADM as input features of NEWFM to diagnose depression. Power_d3 and Power_d2 in particular, having the highest evaluation of the accumulated counters among the 22 initial features, are the most distinguishable between depression and control subjects, which may aid in diagnosing depression and understanding the relationship between HRV and depression.
대상 데이터
This study designed a experiment which used for HRV signals collection. In the experiment, HRV signals were collected from 24 subjects which wore a wireless Holter monitor. Each subject underwent a 13-minute affective-contents stimulus.
This study involved 10 subjects who were diagnosed with depressive disorders (SDS score: greater than 60) at the Mental Health Center, Sungnam, South Korea. Another 14 participants were included as control subjects (SDS score: 20~50).
This study involved 10 subjects who were diagnosed with depressive disorders (SDS score: greater than 60) at the Mental Health Center, Sungnam, South Korea. Another 14 participants were included as control subjects (SDS score: 20~50). They were all healthy volunteers without any history of heart disease or neurological or psychiatric illness.
They were all healthy volunteers without any history of heart disease or neurological or psychiatric illness. The 24 subjects included 9 females and 15 males, none of whom had exercised heavily in the 4 hours prior to taking part in the experiment.
성능/효과
In feature selection process, six features are selected from these features by Non-overlap Area Distribution Measurement (NADM)[8-12] based on neuro-fuzzy networks: VLF, SDNN, PNN100, Power_d3, Power_d2 and SD2. The selected 6 optimal features are compared with the 22 initial features in terms of depression diagnosis accuracy, with results of 95.8% and 91.67%, respectively. This refinement not only reduces the number of the input features but also increases the classification accuracy by selecting the most discriminating features.
[Table 3] presents a comparison between the performance of the 22 initial features and the performance of the 6 optimal features on the 24 subjects' HRV signal sets. The performance results of the initial features were 100% for sensitivity, 85.7% for specificity, 83.3% for productivity, and 91.67% for accuracy rate. The performance results for the optimal features were 100% for sensitivity, 92.
67% for accuracy rate. The performance results for the optimal features were 100% for sensitivity, 92.8% for specificity, 90.9% for productivity, and 95.8% for accuracy rate. A summary of different methods together with their reported results in terms of the accuracy and the number of features is summarized in [Table 4].
참고문헌 (20)
B. Hosseinifard, M. H. Moradi, and R. Rostami, "Classifying Depression Patients and Normal Subjects Using Machine Learning Techniques," Proceeding of Iranian Conference Electrical Engineering, pp.131-134, 2011.
Y. J. Li and F. Y. Fan, "Classification of Schizophrenia and Depression by EEG with ANNs," Proceeding of IEEE Engineering in Medicine and Biology Society, pp.2679-2682, 2005.
I. Kalatzis, N. Piliourasa, E. Ventourasa, C. C. Papageorgioub, A. D. Rabavilas, and D. Cavourasa, "Design and Implementation of an SVM-based Computer Classification System for Discriminating Depressive Patients from Healthy Controls Using the P600 Component of ERP Signals," Computer Methods and Programs in Biomedicine, Vol.75, No.1, pp.11-22, 2004.
A. K. Rostamabad, J. P. Reilly, G. Hasey, H. Bruin, and D. MacCrimmon, "Diagnosis of Psychiatric Disorders Using EEG Data and Employing a Statistical Decision Model," Proceeding of IEEE Engineering in Medicine and Biology Society, pp.4006-4009, 2010.
G. Licht, E. Geus, F. Zitman, W. Hoogendijk, R. Dyck, and B. Penninx, "Association Between Major Depressive Disorder and Heart Rate Variability in the Netherlands Study of Depression and Anxiety," Archives of General Psychiatry, Vol.65, No.12, pp.1358-1367, 2008.
M. Karavidas, "Heart Rate Variability Biofeedback for Major Depression," Applied Psychophysiology and Biofeedback, Vol.36, No.1, pp.18-21, 2008.
M. W. Agelink, C. Boz, H. Ullrich, and J. Andrich, "Relationship between Major Depression and Heart Rate Variability," Clinical Consequences and Implications for Antidepressive Treatment. Vol.113, No.2, pp.139-149, 2002.
J. S. Lim, "Finding Features for Real-Time Premature Ventricular Contraction Detection Using a Fuzzy Neural Network System," IEEE Trans. on Neural, Vol.20, No.3, pp.522-527, 2009.
W. W. Zung, "A Self-rating Depression Scale," Archives of General Psychiatry, Vol.12, No.1, pp.63-70, 1965.
P. S. Hamilton and W. J. Tompkins, "Quantitative Investigation of QRS Detection Rules Using the MIT/BIH Arrhythmia Database," IEEE Trans. on Biomedical Engineering, Vol.33, No.12, pp.1157-1165, 1986.
E. Nahshont, D. Aravot, D. Aizenberg, M. Sigler, G. Zalsman, B. Strasberg, S. Imbra, E. Adler, and A. Weizman, "Heart Rate Variability in Patients With Major Depression," Psychosomatics, Vol.45, No.2, pp.129-134, 2004.
C. K. Lee, S. K. Yoo, Y. J. Park, K. NamHyun, K. S. Jeong, and B. Lee, "Using Neural Network to Recognize Human Emotions from Heart Rate Variability and Skin Resistance," Proceeding of Engineering in Medicine and Biology Society, pp.5523-5525, 2005.
M. Malik, "Heart Rate Variability Standards of Measurement, Physiological Interpretation, and Clinical Use," European Heart Journal, Vol.17, No.3, pp.354-381, 1996.
A. Subasi, "Automatic Recognition of Alertness Level from EEG by Using Neural Network and Wavelet Coefficients," Expert Systems with Applications, Vol.28, No.4, pp.701-711, 2005.
M. Brennan, M. Palaniswami, and P. Kamen "Do Existing Measures of Poincare Plot Geometry Reflect Nonlinear Features of Heart Rate Variability?" IEEE Trans. on Biomedical Engineering, Vol.48, No.11, pp.1342-1347, 2001.
A. Kandaswamy, C. S. Kumar, R. P. Ramanathan, S. Jayaraman, N. Malmurugan, "Neural Classification of Lung Sounds Using Wavelet Coefficients," Computers in Biology and Medicine, Vol.34, No.6, pp.523-537, 2004.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.