Classification Analysis of Occlusion Therapy for Amblyopic (Lazy / Squint eye) using Pattern Visual Evoked Potential (P-VEP).
The term Pattern Visually Evoked Potential (P-VEP) refers to electrical potentials which are recorded from the scalp overlying visual cortex, P-VEP waveforms are extracted from the electro-encephalogram (EEG) by signal averaging.
P-VEP is used primarily to measure the functional integrity of the visual pathways from retina via the optic nerves to the visual cortex of the brain. P-VEPs better quantify functional integrity of the optic pathways than scanning techniques such as magnetic resonance imaging (MRI). The computer programs save a time period of EEG activity following a visual stimulus, which is repeated over and over adding the signals together.
Based on the signal to noise ratio, an evoked potential can be seen forming following only a few stimuli such as flashes of light. Therefore the diagnosis based on amplitude and latency in time domain is not alone sufficient. Hence other components should also be taken into consideration.
The spectral analysis of pattern visual evoked potential can yield useful information when it is performed carefully. Classification of the severity of amblyopic and measuring the changes are vital for assessing the occlusion therapy. Present clinical studies use analysis of visual acuity manually. This method is time-consuming process which requires significant training and exercise and is uncovered to observe error
MATERIALS AND METHODS
In the experiment we used a dataset of EEG signals. The set contains data from 50 subjects acquired from 180 trails. The subjects were relaxed, sat in normal chairs with arms resting on their legs at a distance of 90cm away from monitor.
The experimental task is depicted in Figure. Symbol was flashed on a computer screen, P-VEP waveforms were recorded from 3 electrodes. Reference electrode usually placed on the earlobe, on the midline on top of the head or on the forehead. Ground electrode can be placed at any location.
The time period analyzed is usually between 100 and 500 milliseconds following onset of each visual stimulus. When testing young infants, analysis time was 300 msec or longer because components of the P-VEPs may have long peak latencies during early maturation.
Commonly used visual stimuli are strobe flash, flashing light emitting diodes (LEDs), transient and steady state pattern reversal and pattern onset/offset. The most common stimulus used is a checkerboard pattern, which reverses every half second.
Pattern reversal is a preferred stimulus because there is more inter--subject P-VEP reliability than with flash or pattern onset stimuli. Camera shutters on each projector controlled the display of each checkerboard with red fixation point as shown in figure 2 reversing at a rate 2 per second.
2. Feature Reduction By Linear Discriminant Analysis:
Linear discriminant analysis (LDA) is a well known feature reduction technique . LDA is used to find a linear combination of features that can better separate two or more classes. There are many possibilities for finding directions but only some are optimal for data discrimination. P-VEP signal was extracted from the Oz electrode.
The sample signal is shown in the figure 3. The extracted signal contains the following criteria: Latency criteria, Amplitude criteria, Topographic criteria and waveform criteria. By using Linear Discriminant Analysis algorithm, the features necessary to detect the amblyopia has been identified as Latency and Amplitude criteria with respect to time.
3. Classification Algorithms:
K-means is one of the simplest unsupervised learning algorithms that solve the well-known clustering problem. The procedure follows a simple and easy way to classify a given data set through a certain number of clusters (assume k clusters) fixed a priori. The main idea is to define k centroids, one for each cluster. These centroids should be placed in a cunning way because of different location causes different result. So, the better choice is to place them as much as possible far away from each other.
The next step is to take each point belonging to a given data set and associate it to the nearest centroid. When no point is pending, the first step is completed and an early clustering is done. At this point we need to recalculate k new centroids as berry centers of the clusters resulting from the previous step.
After we have these k new centroids, a new binding has to be done between the same data set points and the nearest new centroid. A loop has been generated. As a result of this loop we may notice that the k centroids change their location step by step until no more changes are done. In other words centroids do not move any more.
Finally, this algorithm aims at minimizing an objective function, in this case a squared error function. The objective function
J = [[SIGMA].sup.k.sub.j=1][[SIGMA].sup.n.sub.i=1][[parallel][x.sup.(j).sub.i]--[c.sub.j][parallel].sup.2]
where [[parallel][x.sup.(j).sub.i]--[c.sub.j][parallel].sup.2] is a chosen distance measure between a data point [x.sup.(j).sub.i] and the cluster centre [c.sub.j] (here 16mv for amplitude and 120 ms for latency criteria respectively), is an indicator of the distance of the n data points from their respective cluster centers.
3.2 Support Vector Machine:
Support vector machines (SVMs) are build on developments in computational learning theory. Because of their accuracy and ability to deal with a large number of predictors, they have more attention in biomedical applications. The majority of the previous classifiers separate classes using hyper planes that split the classes, using a flat plane, within the predictor space. SVMs broaden the concept of hyper plane separation to data that cannot be separated linearly, by mapping the predictors onto a new, higher-dimensional space in which they can be separated linearly. A SVM locates the hyper plane that divides the support vectors without ever representing the space explicitly.
The two classes can only be separated absolutely by a complex curve in the original space of the predictor. The best linear separator cannot totally separate the two classes. On the other hand, if the original predictor values can be projected into a more suitable feature space, it is possible to separate completely the classes with a linear decision boundary.
As a result, the problem becomes one of finding the suitable transformation. The kernel function, which is central to the SVM approach, is also one of the main problems, especially with respect to the selection of its parameter values. It is also crucial to select the magnitude of the penalty for violating the soft margin between the classes.
This means that successful construction of a SVM necessitates some decisions that should be informed by the data to be classified. In contrast, the support vector classifier chooses one particular solution: the classifier which separates the classes with maximal margin. The margin is defined as the width of the largest 'tube' not containing samples that can be drawn around the decision boundary. It can be proven that this particular solution has the highest generalization ability. By using the threshold values from the table. 1 the classification was made.
To calculate the accuracy of the classification Sensitivity and Specificity is to be determined. Sensitivity, also called the true positive ratio, is calculated by the formula:
Sensitivity = TPR = TP/TP+FN x 100%
On the other hand, specificity value (true negative, same diagnosis as the expert neurologists) is calculated by dividing the total of diagnosis numbers to total diagnosis numbers. Specificity, also called the true negative ratio, and accuracy is calculated by the formula TN
Specificity = TNR = TN/TN+FP x 100%
Accuracy = (TP+TN)/(TP+FN+TN+FP)
Test has been carried out of 180 trials for subject S1 and 170 trials for S2 and 150 trials for S3. Since detecting P100 in a single trial is very difficult and therefore repeated stimuli is needed. The classification accuracy is increased by having more number of trials. This was achieved by plotting the sensitivity (defined as the ratio of true positive detected by K-means and SVM algorithm to the total number of true spikes determined by visual analysis) and specificity (defined as the ratio of true positive to the total number of "Neal spikes" detected by K-means and SVM algorithm) as a function of varying thresholds, with the crossover point being regarded as the optimal threshold value.
For age 7 to 12 occlusion therapy will give significant improvement in vision. Compare to Part time & full time therapy full time therapy is better for heavy squint. Part -time therapy is better for mild / moderate squint approximately 6hrs per day.
The pre-processed signals were given as input to two different classification algorithms such as K-means and Support Vector Machine. Using, k-means and Support Vector Machine algorithms the subjects were successfully classified as normal and abnormal. Finally, the performance the classification algorithms were evaluated based upon the Specificity and Sensitivity ratios. Thus the amblyopic has been efficiently classified using SVM algorithm.
[1.] Abdulhamit Subasi a, *, M. Ismail Gursoy (a). International Burch University, Faculty of Engineering and Information Technologies, Sarajevo, Bosnia and Herzegovina
[2.] Kahta Vocational School of Higher Education, Adiyaman University, Adiyaman, Turkey.
[3.] Kantardzic, M., 2002. "Data Mining: Concepts, Models, Methods, and Algorithms", IEEE Press & John Wiley.
[4.] Kantardzic, M., 2002. "Data Mining: Concepts, Models, Methods, and Algorithms", IEEE Press & John Wiley.
[5.] London: Springer. Adeli, H., Z. Zhou and N. Dadmehr, 2003. Analysis of EEG records in an epileptic Patient using wavelet transforms. Journal of Neuroscience Methods, 123: 69-87.
(1) Bairavi. N, (2) Asvedha. B, (3) Aarthi. A, (4) Mrs.Kalaivaazhi Vijayaragavan
(1,2,3,4) Anjalai Ammal Mahalingam Engineering College, Kovilvenni, Thiruvarur
Received 14 September 2017; Accepted 15 October 2017; Available online 30 October 2017
Address For Correspondence:
Bairavi. N, Anjalai Ammal Mahalingam Engineering College,Kovilvenni,Thiruvarur E-mail: firstname.lastname@example.org
Caption: Fig. 1: Overview
Caption: Fig. 2: Checkerboard with Red
Caption: Fig. 3: Sample Signal
Caption: Fig. 4: K-means algorithm
Caption: Fig. 5: SVM Algorithm
Caption: Fig. 6: SVM Normal
Caption: Fig. 7: SVM Abnormal
Caption: Fig. 8: Graphical Representation Of SVM
Table 1: Threshold Values For Clustering Parameters Normal Abnormal Amplitude(mv) 0-16 Above 16 Latency(ms) 0-120 Above 120 Table 2: SVM Statistical S1 (180 S2 (170 S3 (150 parameters Trials) Trials) Trials) TP 50 30 19 FN 3 1 1 TN 31 25 29 FP 4 3 4 Specificity 94% 96% 95% Sensitivity 88% 89% 87% Accuracy 92% 93% 90% Table 3: K-means Statistical S1 (180 S2 (170 S3 (150 parameters Trials) Trials) Trials) TP 47 28 24 FN 3 2 2 TN 28 25 30 FP 4 3 4 Specificity 94% 93% 92% Sensitivity 87% 89% 88% Accuracy 91% 91% 90% Table 4: Classification Accuracy Algorithm Accuracy Specificity Sensitivity SVM 91.5% 95% 88% K-means 90.5% 93% 88% Table 5: Effect of Occlusion Therapy with Duration and Response Rate Based on Age Age Affected Range Duration Treatment Response Rate 7 to 12 Mild/Moderate 2 to 4 hrs 56% 7 to 12 Mild/Moderate 6hrs 76% 7 to 12 Heavy Full Time 90%
|Printer friendly Cite/link Email Feedback|
|Author:||Bairavi, N.; Asvedha, B.; Aarthi, A.; Vijayaragavan, Kalaivaazhi|
|Publication:||Advances in Natural and Applied Sciences|
|Date:||Oct 1, 2017|
|Previous Article:||Effective Fog Networking Communications using An Attribute-Based Encryption Scheme Decentralized Server.|
|Next Article:||Preserving security and privacy in mobile cloud computing health care service w.r.t noncontact ECG monitoring.|