Home

Intelligent methods for non-invasive computer-aided diagnosis of laryngopathies

       The research was supported by the grant No. N N516 423938 from the Polish Ministry of Science and Higher Education.

Research Team:

Krzysztof Pancerz (Univerisity of Information Technology and Management in Rzeszow) – scientific leader
Prof. Jerzy Grzymała-Busse (University of Kansas, Univerisity of Information Technology and Management in Rzeszow) – expert in data mining and knowledge discovery
Jarosław Szkoła (Univerisity of Information Technology and Management in Rzeszow)
Jan Warchoł (Medical Univeristy of Lublin)
Maksymilian Knap (Univerisity of Information Technology and Management in Rzeszow)
Wiesław Paja (Univerisity of Information Technology and Management in Rzeszow)


Research Results:

       The aim of the project was to develop intelligent methods and algorithms supporting a non-invasive diagnosis of selected larynx diseases and to implement them in a dedicated (specialized) software tool. Two larynx diseases, Reinke's edema and laryngeal polyp, have been taken into consideration on the basis of available samples of voice signals. The diagnosis has been based on analysis of selected parameters of voice signals in both a time domain and a frequency domain. In the time domain, an application of recurrent neural networks (modified Elman-Jordan networks) has been proposed. It was a new neural network model which allows to undertake a learning process in the efficient way. Moreover, approaches using rough set theory (calculation of the so-called consistency factors) as well as ant-based clustering have been proposed. In the time domain, an original signal and its derivative have been analyzed. Additionally, approximation of the original signal using Bezier curves has been introduced. In this case, a neural network was trained on parameters of approximation curves instead of signal samples. A special iterative algorithm for finding a family of Bezier curves best approximating a given signal has been developed. In order to approximate a voice signal, 4-point Bezier curves have been used. Bezier curve approximation reduced amount of data to be learned as well as removed a noise from the original signal. In the frequency domain, a family of coefficients reflecting spectrum disturbances around basic tones and their multiples has been proposed. Observations showed that distribution of spectrums of voice signals for patients without diseases was more regular. Moreover, it was easy to see more distinct slenderness of the distribution around basic tones and their multiples. The proposed family of coefficients expressed slenderness of the spectrum in regions of interest as well as regularity of distribution of basic tones and their multiples. Description of cases by calculated coefficients was a base for a classification process. In this process, a variety of data mining and machine learning approaches has been applied. The following classification algorithms implemented in the available computer tools have been used: WEKA (decision tree algorithms), RSES (rule based algorithms), NGTS (rule based algorithm), BeliefSEEKER (algorithm based on belief networks). Moreover, optimization and verification of rules knowledge bases have been performed using methods implemented in RuleSEEKER. Belief networks has been also used in the attribute selection process. For the majority of approaches, a classification accuracy greater than 80% has been obtained. It was a satisfactory result due to the character of available data.
       Prototype environments have been created for developed methods and algorithms. In case of time domain analysis, the MATLAB system has been used due to availability of tools for creation and simulation of models in the form of neural networks (Neural Network Toolbox). Methods and algorithms based on frequency domain analysis have been implemented in two environments, Scilab and LabVIEW. In the second case, LabVIEW allows creating, in easy way, a user-friendly graphical interface facilitating both entering input data and visualizing results in order to make the platform ready to use directly in the medical community. Mentioned environments have been used for verification of the developed methods and algorithms on real-life data. Voice samples for two groups have been used for testing them (a control group as well as a group of patients from Otolaryngology Clinic of the Medical University of Lublin in Poland).
       A final result of the project is a computer tool created in the Java language and implementing developed approaches supporting a diagnosis of selected larynx diseases. The tool is called LARDISS (the acronym comes from LARyngopathy DIagnosis Support System). Beta versions of the developed tolls are available in Section “Software tools”.


Publications:

• Warchoł, J., Szkoła, J., Pancerz, K.: Towards Computer Diagnosis of Laryngopathies Based on Speech Spectrum Analysis: A Preliminary Approach. In: A. Fred, J. Filipe, H. Gamboa (Eds.), Proceedings of the Third International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS'2010), Valencia, Spain, January 20-23, 2010, pp. 464-467. Abstract

• Szkoła, J., Pancerz, K., Warchoł, J.: Computer Diagnosis of Laryngopathies Based on Temporal Pattern Recognition in Speech Signal. Bio-Algorithms and Med-Systems, Vol. 6, No. 12, 2010,
pp. 75-80. Abstract

• Szkoła, J., Pancerz, K, Warchoł. J.: Computer-Based Clinical Decision Support for Laryngopathies Using Recurrent Neural Networks. In: A.E. Hassanien, A. Abraham, F. Marcelloni, H. Hagras, M. Antonelli, T.-P. Hong (Eds.), Proceedings of the The 10th International Conference on Intelligent Systems Design and Applications (ISDA’2010), Cairo, Egypt, November 29 – December 1, 2010, pp. 627-632. Abstract

• Szkoła, J., Pancerz, K., Warchoł, J.: Improving Learning Ability of Recurrent Neural Networks: Experiments on Speech Signals of Patients with Laryngopathies. In: F. Babiloni, A. Fred, J. Filipe, H. Gamboa (Eds.), Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS'2011), Rome, Italy, January 26-29, 2011, pp. 360-364. Abstract

• Szkoła, J., Pancerz, K., Warchoł, J.: Recurrent Neural Networks in Computer-Based Clinical Decision Support for Laryngopathies: An Experimental Study. Computational Intelligence and Neuroscience, Vol. 2011, Article ID 289398, 2011. Abstract    Full paper

• Pancerz, K., Szkoła, J., Warchoł, J.: A Bezier Curve Approximation of the Speech Signal in the Classification Process of Laryngopathies. In: M. Ganzha, L. Maciaszek, M. Paprzycki (Eds.), Proceedings of the Federated Conference on Computer Science and Information Systems (FedCSIS'2011), Szczecin, Poland, September 18-21, 2011, pp. 141-146. Abstract

• Pancerz, K., Szkoła, J., Warchoł, J., Olchowik G.: Spectrum Disturbance Analysis For Computer-Aided Diagnosis of Laryngopathies: An Exemplary Study. Proceedings of the International Workshop on Biomedical Informatics and Biometric Technologies (BT'2011), Zilina, Slovak Republic, November 10-11, 2011. Abstract

• Pancerz, K., Paja, W., Szkoła, J., Warchoł, J., Olchowik, G.: A Rule-Based Classification of Laryngopathies Based on Spectrum Disturbance Analysis - An Exemplary Study. In: S. Van Huffel, C. Correia, A. Fred, H. Gamboa (Eds.), Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS'2012), Vilamoura, Algarve, Portugal, February 1-4, 2012, pp. 458-461. Abstract

• Mroczek, T., Pancerz, K., Warchoł, J.: Belief Networks in Classification of Laryngopathies Based on Speech Spectrum Analysis. In: T. Li et al. (Eds.), Proceedings of the 7th International Conference on Rough Sets and Knowledge Technology (RSKT'2012), Chengdu, China, August 17-20, 2012, Lecture Notes in Artificial Intelligence, Vol. 7414, Springer-Verlag, Berlin Heidelberg, 2012, pp. 222-231. Abstract

• Szkoła, J., Pancerz, K., Warchoł, J., Olchowik, G., Klatka, M., Wojecka-Gieroba, R., Wróbel, A.: LARDISS - a Tool for Computer Aided Diagnosis of Laryngopathies. In: M. Ganzha, L. Maciaszek, M. Paprzycki (Eds.), Proceedings of the Federated Conference on Computer Science and Information Systems (FedCSIS'2012), Wroclaw, Poland, September 9-12, 2012, pp. 227-233. Abstract

• Pancerz, K., Paja, W., Wrzesień, M., Warchoł, J.: Classification of Voice Signals through Mining Unique Episodes in Temporal Information Systems: A Rough Set Approach. In: L. Popova-Zeugmann (Ed.), Proceedings of the Workshop on Concurrency, Specification and Programming (CS&P'2012), Berlin, Germany, September 26-28, 2012, Vol. 2, pp. 280-291.

• Gurdak, D., Pancerz, K., Szkola, J., Warchol, J.: Computer-Aided Diagnosis of Laryngopathies in the LabVIEW Environment: Exemplary Implementation. In: Kountchev, R., Iantovics, B. (Eds.), Advances in Intelligent Analysis of Medical Data and Decision Support Systems. Studies in Computational Intelligence, Vol. 473, Springer-Verlag, Berlin Heidelberg, 2013, pp. 157-167.