Ensemble adaptive network-based fuzzy inference system with weighted arithmetical mean and application to diagnosis of optic nerve disease from visual-evoked potential signals

https://doi.org/10.1016/j.artmed.2008.03.007Get rights and content

Summary

Objective

This paper presents a new method based on combining principal component analysis (PCA) and adaptive network-based fuzzy inference system (ANFIS) to diagnose the optic nerve disease from visual-evoked potential (VEP) signals. The aim of this study is to improve the classification accuracy of ANFIS classifier on diagnosis of optic nerve disease from VEP signals. With this aim, a new classifier ensemble based on ANFIS and PCA is proposed.

Methods and material

The VEP signals dataset include 61 healthy subjects and 68 patients suffered from optic nerve disease. First of all, the dimension of VEP signals dataset with 63 features has been reduced to 4 features using PCA. After applying PCA, ANFIS trained using three different training–testing datasets randomly with 50–50% training–testing partition.

Results

The obtained classification results from ANFIS trained separately with three different training–testing datasets are 96.87%, 98.43%, and 98.43%, respectively. And then the results of ANFIS trained with three different training–testing datasets randomly with 50–50% training–testing partition have been combined with three different ways including weighted arithmetical mean that proposed firstly by us, arithmetical mean, and geometrical mean. The classification results of ANFIS combined with three different ways are 98.43%, 100%, and 100%, respectively. Also, ensemble ANFIS has been compared with ANN ensemble. ANN ensemble obtained 98.43%, 100%, and 100% prediction accuracy with three different ways including arithmetical mean, geometrical mean and weighted arithmetical mean.

Conclusion

These results have shown that the proposed classifier ensemble approach based on ANFIS trained with different train–test datasets and PCA has produced very promising results in the diagnosis of optic nerve disease from VEP signals.

Introduction

Medical decision making systems including data mining tools, machine learning algorithms, data pre-processing methods are often used in diagnosing any disorder by means of signal or image data obtained from patient in the area of medical diagnosis. There is absolutely that evaluation of data taken from patient and decisions of experts are the most important factors in diagnosis. But, expert systems and different artificial intelligence techniques for classification help also experts in a great deal [1].

Optic nerve is generally taken into consideration a component of the central nervous system. Damage to optic nerve fibers can happen at or near their origin in the retina or in the nerve, optic tract, or lateral geniculate nuclei. Clinical manifestations can contain decreased visual acuity and contrast sensitivity and an afferent papillary deficiency. The pattern visual evoked potential (VEP) is greatly recognized as a sensitive measure of optic nerve pathologies, including demyelization and is a massed cortical response elicited by a change in retinal stimulation primarily driven by the fovea [2].

The VEP signals connected to retinal functions and optics behaviors. Thus VEP signals can be used to diagnose optic nerve diseases. VEP signals are being used in clinical tests and consultation.

The VEP is an evoked electrophysiological potential that can be extracted, using signal averaging from the electroencephalographic activity recorded at the scalp. It represents mass activity of all cells in the visual cortex. A complex wave is generated with discernible positive and negative peaks that occur at predictable latency times after the visual stimulus [3]. The VEP can provide important diagnostic information regarding the functional integrity of the visual system and it has been used in clinical and research laboratories. The VEP can be abnormal in diseases of the outer retina such as optic nerve disease or hereditary macular degeneration, as well as in diseases of the optic nerve or visual cortex. In cases of poor vision without evident retinal disease, VEP can be used to probe the integrity of the optic nerve and cortical tracts. The VEP is more useful in evaluating cases of reduced visual acuity versus constricted visual fields [4], [5]. In general, the clinical use of VEP is based on the peak amplitude and the latencies of the N75, P100, and N145 waves (Figure 1, Figure 2). If there are ophthalmologic disorders, VEP recordings change in latency and the diagnosis is based on the measurement of latency directly from the signal. In certain cases, background electroencephalogram (EEG) found to have effect on VEP wave-forms, which in turn results in irregular peaks and special processing techniques like averaging and interpolation have to be done to overcome these disorders [5], [6], [7].

A physician usually makes decisions by evaluating the current test results of a patient and by referring to the previous decisions he/she made on other patience with the same condition. The former method depends strongly on the physician's knowledge. On the other hand, the latter depends on the physician's experience to compare her/his patient with her/his earlier patients. This job is not easy considering the number of factors she/he has to evaluate. In this crucial step, she/he may need an accurate tool that lists her/his previous decisions on the patient having same (or close to same) factors [8].

The proposed system has four stages: (i) feature reduction using PCA, (ii) production of different VEP datasets by random, (iii) training of ANFIS using these datasets, and (iv) improving the classification accuracy of ANFIS using three different ways including weighted arithmetical mean that was proposed firstly by us, arithmetical mean, and geometrical mean.

The remaining of the paper is organized as follows. The data acquisition is presented in the next section. The proposed method is explained in Section 3. The experimental data to show the effectiveness of our method is given in Section 4. Finally, Section 5 presents the discussion and future directions.

Section snippets

Data acquisition: VEP signals

In this study, VEP signals experiments were carried out with 129 subjects. The group consists of 55 females and 74 males with ages ranging from 33 to 49 years and mean age of 43.5 years (standard deviation, S.D. = 4.9). Electrophysiological test devices were used during examinations and signals were taken into consideration. According to examination results, 61 of 129 subjects were healthy subjects and the rest of them were optic nerve diseased subjects (Fig. 2). The group having optic nerve

Overview

The proposed method consists of three parts. In the first stage, PCA applied to VEP signals with 63 features and the dimension of VEP signals was reduced from 63 to 4 features via PCA. In the second stage, ANFIS trained on the three random data sets by means of VEP signals that have four features. Finally the predictions of the ANFIS combined with three ways including weighted arithmetical mean that is firstly proposed by us, arithmetical mean, and geometrical mean. In the arithmetical mean

The empirical results and discussion

To evaluate the effectiveness of our method, the experiments on the optic nerve disease database mentioned above were made. In this study, the principal component analysis (PCA) and adaptive network-based fuzzy inference system (ANFIS) to diagnose the optic nerve disease from VEP signals were combined. The aim of this study was to improve the classification accuracy of ANFIS classifier on diagnosis of optic nerve disease from VEP signals. With this aim, a new classifier ensemble based on ANFIS

Conclusion

In this paper, a novel classifier ensemble method called weighted arithmetical mean to diagnose optic nerve disease from VEP signals was proposed. This method was combined with ANFIS and ANN classifier and compared to other combination methods including arithmetical mean and geometrical mean. Ensemble ANFIS outperforms ensemble ANN in diagnosis of optic nerve disease. VEP signals are used to diagnose both optic nerve disease and macular disease. Ensemble ANFIS to diagnose both optic nerve

Acknowledgment

This study is supported by the Scientific Research Projects of Selcuk University.

References (18)

There are more references available in the full text version of this article.

Cited by (0)

View full text