Automatic Diagnosis of Parkinson's Disease by Applying ICA Methods to TQWT Features

Year 2019, Volume: 6, 50 - 58, 30.09.2019

Cüneyt Yücelbaş , Şule Yücelbaş

https://doi.org/10.35193/bseufbd.566857

Cited By: 2

Abstract

Parkinson is a
disease caused by the loss of dopamine-producing brain cells. There are many
diagnostic methods of this disease, and the analysis of audio signals is one of
them. For this purpose, the features obtained by using the tunable Q-factor
wavelet transform (TQWT) method were used for the recorded audio signals of a
total of 252 people including 188 Parkinson's disease and 64 healthy subjects.
Three different feature reduction (dimensionality reduction) methods as fast
ICA (FICA), max-kurtosis ICA (KICA) and reconstruction ICA (RICA) which are one
of the independent component analysis (ICA) were applied to these properties.
As a result of these processes, the maximum success rate was tried to be
obtained with the minimum number of features. For this purpose, firstly k-fold
cross validation method is applied to the data group created with new features
and the data are divided into train-test. In the next step, the prepared data
were classified by Random Forest (RO) algorithm and the results were
interpreted by various statistical criteria. When the results are evaluated;
the most successful method was the RICA with 82.01 classification accuracy and
the ROC and PRC values of about 0.85. This situation has proved almost perfect
separation of the patient and the healthy class. The performance results of
this study which is suitable for real life applications and the high number of
data used reveal the importance of the study in the literature. Moreover, the
analysis of dimensionality reduction methods used in the study can lead to
studies that can be done in this area. 

Keywords

Parkinson's disease, audio signal analysis, dimensionality reduction methods, TQWT

AQDD Özelliklerine BBA Yöntemleri Uygulanarak Parkinson Hastalığının Otomatik Teşhisi

Abstract

Parkinson hastalığı dopamin
üreten beyin hücrelerinin kaybı sonucunda oluşan bir hastalıktır. Bu hastalığın
birçok teşhis yöntemi bulunmakta olup ses sinyallerinin analizi de bunlardan
birisidir. Bu çalışmada daha önceden 188 Parkinson hastası ve 64 sağlıklı olmak
üzere toplam 252 kişiye ait kaydedilmiş ses sinyallerinden ayarlanabilir
Q-faktör dalgacık dönüşümü (AQDD) metodu kullanılarak elde edilen özellikler
kullanılmıştır. Bu özelliklere bağımsız bileşen analizi (BBA) çeşitlerinden
olan hızlı BBA (HBBA), max-kurtosis BBA (KBBA) ve yeniden yapılanma BBA (YBBA)
olmak üzere üç farklı özellik azaltma (boyut indirgeme) yöntemi uygulanmıştır.
Bu işlemler sonucunda minimum özellik sayısıyla maksimum başarı oranı elde
edilmeye çalışılmıştır. Bu amaçla, öncelikle yeni özellikler ile oluşturulan
veri grubuna ayrı ayrı k-kat çapraz doğrulama yöntemi uygulanarak veriler
eğitim-test olarak ayrılmıştır. Sonraki aşamada, hazırlanan veriler rastgele orman
(RO) algoritması ile sınıflandırılmış ve sonuçlar çeşitli istatistiksel
ölçütlerle yorumlanmıştır. Sonuçlar değerlendirildiğinde; kullanılan boyut
indirgeme yöntemleri içerisinde en başarılı yöntem %82.01 sınıflandırma
doğruluk oranı ve yaklaşık 0.85 ROC ve PRC değerleri ile YBBA olmuştur. Bu
durum hasta ve sağlıklı sınıf ayrışımının mükemmele yaklaştığını kanıtlamıştır.
Gerçek yaşam uygulamalarına uygun olan bu çalışmanın performans sonuçları ve
kullanılan veri sayısının yüksek oluşu çalışmanın literatürdeki önemini ortaya
koymaktadır. Ayrıca, çalışma kapsamında kullanılan özellik indirgeme
yöntemlerinin analizi, bu alanda yapılabilecek çalışmalara yol gösterebilecek
niteliktedir.

Keywords

Parkinson hastalığı, ses sinyali analizi, boyut indirgeme metodları, AQDD

References

• [1] K. Rana. (2014, 26 December 2018). Parkinson Hastalığı [Online]. Available: http://www.noroloji.org.tr/TNDData/Uploads/files/parkinson%20hastal%C4%B1%C4%9F%C4%B1.pdf.
• [2] S. Özekmekçi, H. Apaydın, S. Oğuz, & İ. Zileli. (2013). Parkinson Hastalığı Hasta ve Yakınları İçin El Kitabı. İstanbul, Turkey: Bayçınar Tıbbi Yayıncılık ve Reklam Hiz. Tic. Ltd. Şti, p. 98.
• [3] J. W. Langston. (2002). Parkinson’s disease: current and future challenges. Neurotoxicology, vol. 23, no. 4-5, pp. 443-450.
• [4] J. Parkinson. (1817). An essay on the shaking palsy (Printed by Whittingham and Rowland for Sherwood, Neely, and Jones), ed: London.
• [5] J. Jankovic. 2008). Parkinson’s disease: clinical features and diagnosis. Journal of neurology, neurosurgery & psychiatry, vol. 79, no. 4, pp. 368-376.
• [6] H. Gümüş, Z. Akpınar, & O. Demir. (2013). Erken evre Parkinson hastalığında motor olmayan semptomların değerlendirilmesi. Türk Nöroloji Dergisi, vol. 19, no. 3, pp. 97-103.
• [7] Y. Akgün & S. Peker. (2010). Tremor tedavisinde cerrahi girişimler. Acıbadem Üniversitesi Sağlık Bilimleri Dergisi, vol. 1 (3), no. 3, pp. 123-127.
• [8] B. Harel, M. Cannizzaro, & P. J. Snyder. (2004). Variability in fundamental frequency during speech in prodromal and incipient Parkinson's disease: A longitudinal case study. Brain and cognition, vol. 56, no. 1, pp. 24-29.
• [9] A. Tsanas, M. A. Little, P. E. McSharry, & L. O. Ramig. (2010). Accurate telemonitoring of Parkinson's disease progression by noninvasive speech tests. IEEE transactions on Biomedical Engineering, vol. 57, no. 4, pp. 884-893.
• [10] C. O. Sakar & O. Kursun. (2010). Telediagnosis of Parkinson’s disease using measurements of dysphonia. Journal of medical systems, vol. 34, no. 4, pp. 591-599.
• [11] H. Gürüler. (2017). A novel diagnosis system for Parkinson’s disease using complex-valued artificial neural network with k-means clustering feature weighting method. Neural Computing and Applications, vol. 28, no. 7, pp. 1657-1666.
• [12] M. A. Little, P. E. McSharry, E. J. Hunter, J. Spielman, & L. O. Ramig. (2009). Suitability of dysphonia measurements for telemonitoring of Parkinson's disease. IEEE transactions on biomedical engineering, vol. 56, no. 4, pp. 1015-1022.
• [13] M. Peker, B. Sen, & D. Delen. (2015). Computer-aided diagnosis of Parkinson’s disease using complex-valued neural networks and mRMR feature selection algorithm. Journal of healthcare engineering, vol. 6, no. 3, pp. 281-302.
• [14] B. E. Sakar ve ark. (2013). Collection and analysis of a Parkinson speech dataset with multiple types of sound recordings. IEEE Journal of Biomedical and Health Informatics, vol. 17, no. 4, pp. 828-834.
• [15] R. Das. 2010). A comparison of multiple classification methods for diagnosis of Parkinson disease. Expert Systems with Applications, vol. 37, no. 2, pp. 1568-1572.
• [16] F. Åström & R. Koker. 2011). A parallel neural network approach to prediction of Parkinson’s Disease. Expert systems with applications, vol. 38, no. 10, pp. 12470-12474.
• [17] Ö. Eskidere, F. Ertaş, & C. Hanilçi. (2012). A comparison of regression methods for remote tracking of Parkinson’s disease progression. Expert Systems with Applications, vol. 39, no. 5, pp. 5523-5528.
• [18] B. E. Sakar, G. Serbes, & C. O. Sakar. (2017). Analyzing the effectiveness of vocal features in early telediagnosis of Parkinson's disease. PloS one, vol. 12, no. 8, p. e0182428.
• [19] D. Braga, A. M. Madureira, L. Coelho, & R. Ajith. (2019). Automatic detection of Parkinson’s disease based on acoustic analysis of speech. Engineering Applications of Artificial Intelligence, vol. 77, pp. 148-158.
• [20] C. O. Sakar ve ark. (2019). A comparative analysis of speech signal processing algorithms for Parkinson’s disease classification and the use of the tunable Q-factor wavelet transform. Applied Soft Computing, vol. 74, pp. 255-263.
• [21] Ö. Eskidere. (2012). A Comparison Of Feature Selection Methods For Diagnosis Of Parkinson’s Disease From Vocal Measurements. Sigma, vol. 30, pp. 402-414.
• [22] I. W. Selesnick. (2011). Wavelet transform with tunable Q-factor. IEEE transactions on signal processing, vol. 59, no. 8, pp. 3560-3575.
• [23] D. P. Acharya, G. Panda, & Y. Lakshmi. (2010). Effects of finite register length on fast ICA, bacterial foraging optimization based ICA and constrained genetic algorithm based ICA algorithm. Digital Signal Processing, vol. 20, no. 3, pp. 964-975.
• [24] S. Jiang, P. Lin, Y. Chen, C. Tian, & Y. Li. (2019). Mixed-signal extraction and recognition of wind turbine blade multiple-area damage based on improved Fast-ICA. Optik, vol. 179, pp. 1152-1159.
• [25] V. Zarzoso & P. Comon. (2010). Robust independent component analysis by iterative maximization of the kurtosis contrast with algebraic optimal step size. IEEE Transactions on Neural Networks, vol. 21, no. 2, pp. 248-261.
• [26] R. Xixi & Z. Qun. (2011). Power quality harmonic detection based on Fast-ICA. in 2011 IEEE Power Engineering and Automation Conference, vol. 3: IEEE, pp. 26-29.
• [27] T. Ahmad & M. Ghanbari. (2011). A review of independent component analysis (ica) based on kurtosis contrast function. Australian Journal of Basic and Applied Sciences, vol. 5, no. 9, pp. 1747-1755.
• [28] J. Wang, C. Wang, T. Zhang, & B. Zhong. (2016). Comparison of different independent component analysis algorithms for output-only modal analysis. Shock and Vibration, vol. 2016.
• [29] H. Li & T. Adali. (2008). A class of complex ICA algorithms based on the kurtosis cost function. IEEE Transactions on Neural Networks, vol. 19, no. 3, pp. 408-420.
• [30] A. Hyvärinen, J. Karhunen, & E. Oja. (2004). Independent component analysis. John Wiley & Sons.
• [31] Q. V. Le, A. Karpenko, J. Ngiam, & A. Y. Ng. (2011). ICA with reconstruction cost for efficient overcomplete feature learning. in Advances in neural information processing systems, pp. 1017-1025.
• [32] L. Breiman. (2001). Random forests. Machine learning, vol. 45, no. 1, pp. 5-32.
• [33] I. H. Witten, E. Frank, L. E. Trigg, M. A. Hall, G. Holmes, & S. J. Cunningham. (1999). Weka: Practical machine learning tools and techniques with Java implementations.
• [34] Y. Ma, L. Guo, & B. Cukic. (2006). A statistical framework for the prediction of fault-proneness. Advances in Machine Learning Application in Software Engineering, Idea Group Inc, pp. 237-265.
• [35] D. M. Powers. (2011). Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation.
• [36] N. Nicolov. (2012, 10 April). Machine Learning with Applications in Categorization, Popularity and Sequence Labeling: 57th and 58nd slides. [Online]. Available: http://www.slideshare.net/Nicolas_Nicolov/machine-learning-14528792.
• [37] Ş. Yücelbaş, C. Yücelbaş, G. Tezel, S. Özşen, & Ş. Yosunkaya. (2018). Automatic sleep staging based on SVD, VMD, HHT and morphological features of single-lead ECG signal. Expert Systems with Applications, vol. 102, pp. 193-206.
• [38] T. Saito & M. Rehmsmeier. (2015). The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PloS one, vol. 10, no. 3, p. e0118432.
• [39] T. J. Wroge, Y. Özkanca, C. Demiroglu, D. Si, D. C. Atkins, & R. H. Ghomi. (2018). Parkinson’s Disease Diagnosis Using Machine Learning and Voice. in 2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB): IEEE, pp. 1-7.