Abnormality Segmentation and Detection in Brain Images using Artificial Neural Networks


Call for Papers Engineering Research Journal June 2019

Download Full-Text PDF Cite this Publication

Text Only Version

Abnormality Segmentation and Detection in Brain Images using Artificial Neural Networks

1R. Thillaikkarasi

Assistant Professor,

Department of ECE, Salem College of Engineering and Technology,

Salem – 636 111.

2 S. Saravanan

Assistant Professor,

Department of ECE, Muthayammal Engineering and Coollege,

Salem

3 A. Chitra

Post Graduate student, ME in Applied Electronics, Department of ECE, Salem College of Engineering and Technology,

Salem – 636 111.

Abstract:- Abnormality segmentation and detection of brain images is a challenging problem which has received much attention during the recent years due to its vital importance in medical fields. Semi supervised learning does not require pathology modeling and thus allows high degree of automation. The methodology for abnormality segmentations uses a set of pathology free images in order to calculate on objective function measuring similarity to a healthy brain and a test image for which the objective function is maximized. The above mentioned method is applied for segmenting brain pathologies such as simulated brain infarction and dysplasia as well as real lesions in diabetic patients. However since abnormalities are usually rare or there may even be no data that describe specific pathologic conditions, also it requires substantial human effort and is often prohibitively expensive. Semi supervised abnormality detection offers a solution to this problem by modeling normal data and then using a distance measure and thresholding to determine abnormality. So the present approach using SPM (Statistical Parametric Mapping) goes beyond the standard anomaly detection techniques in that it not only characterizes the data vector normal or abnormal but also locates which part of the vector includes the anomaly.

I INTRODUCTION

Nowdays most of the people affected in brain Tumor or Brain cancer by using this method easily identify the brain tumor. The basic goal in segmentation process is to partition an image into regions that are homogeneous with respect to one or more characteristics. A new feature vector x is then synthesized, aiming to be similar to x but having the anomalies removed. A spatial abnormality scoremap is subsequently created by voxel-wise subtraction, |x x|. Thresholding of such score map gives the segmentation of image anomalies. In contrast to most brain lesion segmentation methods based on outlier detection , the proposed method is generic.

In order to deal with the large dimensionality, partition

the images into subspaces, i.e., locally coherent overlapping blocks. It is assumed that for each location the blocks are generated from a Gaussian distribution and located in a tight cluster. These subspaces are then modeled by a linear method, such as principal component analysis (PCA).

In 2009 Tao Wang presented to ability to capture large range and extract concave shapes and it demonstrate improvements over technique like gradient vector flow, boundary vector flow and magneto static active on three steps. In 2012 A.Lakshmi presented to detection of range and shape tumor in brain MR images and it performs very fast and simple. High accuracy especially for noise. In this paper is proposed to the method are based on the assumption that normal and abnormal instances can be easier separated in a lower dimensional.

  1. METHODS

    It describes the definition of the local objec- tive function, which consists of five terms (E1 , E2 , E3, E4, E5 ).

    A.Objective Function

    Since anomalies are defined as points with low probability density, it is expected to estimate x by maximizing the pdf obtained for the normal data. However, if the vector is high dimensional, the estimation of the pdf is not feasible.

    Therefore, maximize the pdf in a lower dimensional space p(u), where u is the representation of x in a basis W:

    u = W T x. (1)

    Here, x is a column vector assumed to be centered at the origin and T denotes the matrix transpose. If the Karhunen- Loève (KL) transform (or PCA) is applied, the basis W is

    formed by the (d × d) matrix of the eigenvectors of the covariance matrix C of the training set V, i.e., C = (1/n 1)V T V . The KL transform can be inverted as follows: x = Wu. Assuming that x follows a multivariate Gaussian distribution, the density of u is the multivariate Gaussian density:

    p(u) = 1

    —————— e-(1/2)utD-1 u (2)

    (2)k / 2 |D|1 / 2

    where D = W T CW = diag(1 , 2 , . . . , d ) is a (d × d) diagonal matrix of eigenvalues, assumed to be sorted in descending order. Typically, the number of samples is significantly smaller than the dimensionality in which case the eigenvalues t , with t n, are zero and the corresponding eigenvectors in W are ignored. If all other eigenvectors are retained, u Rn 1 . p(u) is maximized when (1/2)uT D1 u is minimized. Based on maximization of the density in respect to x then is equivalent to minimizing the following term:

    E1(x)= 1/2(xT(WD1WT)x). (3)

    Since u is lower dimensional than x, there exist an infinite num-ber of data points x Rd with the same function cost value in.

    The simplest way to smooth a times series is to calculate a simple, or unweighted, moving average, the smoothed statistics st is then just the mean of the last k observation

    k-1

    St = 1/k xt-n (4)

    N=0

    The smoothening factor applied to here is smoothening of misnomer a larger values of x actually reduce the level of smoothening and limiting case with x=1 the output series in just same as the original series.

    St = E2(x;x0) = x x0-1 + (1-x) st-1 (5)

    The gradient is the change in gray level with direction this can be calculated by taking the difference in value of neighbouring pixels. Let us construct the a new array B that contains the value of gradient from x.

    The horizontal gradient is formed by taking the difference between column value

    Y(j,k) = x(j,k+1) x(j,k)

    Horizontal edges would be detected by calculating the vertical gradient the equation for the separated vertical difference is

    Y(j,k) = x(j+1, k) x(j-1, k)

    If x has gray values in the range 0 to 255.. if the example then may have value in the range -255 to 255.

    y(j,k) – ymin

    ——————————- × 255 ——> j(j,k) (6)

    ymax – ymin

    E3 (x; x0 ) = x(j,K+1) X0(j, k) (7)

    In order to reduce the solution space, we use an additional term that constraints the solution to remain close to the subspace spanned by W. If the test vector is x0 , then its projection on W is

    x0 W = W u = W WT x0 .

    The fourth energy term expresses the distance to the pro- jected point x0 W :

    E4 (x; x0 ) = ||x x0 W ||2 = (x W W T x0 )T (x W W T x0 ) (8)

    where · denotes the L2 -norm. If x = x0 , this term expresses the reconstruction error or residual .Since x0 does not necessarily lie within the subspace spanned by W, this term is larger than zero in this setting. This happens mainly because the abnormal vector x0 is inconsistent with the normal data building the basis W.

    The first two terms statistically model normality and are used to make the image look like if abnormality were removed. The final term is used to constrain the reconstructed image to be as similar as possible to the original image x0 based on the assumption that the majority of the voxels in the test image are normal. If all voxels are equally possible to be abnormal, then the distance from x0 can be used as dissimilarity criterion

    d

    E5 (x; x0 ) = ||x x0 ||2 = (x(j) x0 (j))2

    (9)

    j=1/p>

    where j indicates the voxels in the image.If prior knowledge exists on spatial locations of possible ab- normality, then weights can be incorporated to penalize less the dissimilarity in those locations.

    E5(x;x0)=(xx0)TA(xx0) (10)

    whereA is a (d × d) diagonal matrix with normalized elements a(j)/j=1a(j) on the main diagonal. The uncertainty vector a is calculated as the average reconstruction error at each lo- cation over all training images obtained by leave-one-out cross validation where Wt is the basis formed without using training image t.

    The previous five terms are combined into a single objective function, l(x) by using different weights, shown as follows:

    x = arg min l(x), where

    x

    l(x) = w1 E1 (x) + w2 E2 (x; x0 ) + w3 E3 (x; x0 ) + w4 E4 (x; x0 ) + w5 E5 (x; x0 ) (11)

    and 0w1,w2,w3, w4, w51 and w1+w2+w3 +w4+ w5=1.

    According to the values of the weights, we balance between the model term (including E1 and E4 ), controlling the similar-ity with the training set consisting of normal data, and the data term (E3 ), controlling the similarity with the original vector. The weights depend on the confidence we have on the statisti-cal model, as well as on the dominance of novelty or anomaly over the data. Once the optimization problem is solved, the final recon- structed image is created by recentering to the original space, i.e., by adding the mean image to the result.

  2. RESULTS

    The calculated spatial abnormality score map, |x0 x|, is compared against the region of simulated pathology for the simulated data and the expert-defined lesion mask for the real data. By thresholding the abnormality score map, binary segmentation can be obtained. The results are assessed using ROC performance curves, created by changing the threshold level from 1 to 0. The area under the curve (AUC) gives an overall measure of sensitivity versus specificity and is used as indepen-dent evaluation criterion.

    A. Experimental Result

    The selection of a single scan was performed to reduce the amount of data to process and focus on scans that in-clude pathology. The best performance was obtained for block size 42 × 42 voxels and weighting values w1 = 0.98, w2

    = 0.01, w3 = 0.98, , w3 = 0.98, , w3 = 0.01. The AUC value

    was 0.986. Here, global struc-ture seems to be represented quite satisfactorily by the model likelihood term E1 , whereas local structure variability is not sufficiently well captured using this small number of available training samples. This is attributed mainly to the preprocessing of the MR images that included registration by a deformable transformation. These experiments show that the results cannot be generalized and that the optimal weights are not the same for any kind of data. Thus, we decided to equally weight all terms using w1 = w2 = w3 = w4 = w5 = 1/5. We achieved AUC = 0.981 and 0.973 for the simulated and real data, respectively.

    The average (over all subjects) AUC value obtained by the voxel-wise standard score was significantly smaller (0.905), which indicates better performance of the current approach com-pared to univariate statistics.

    Fig 1. Assessment of the proposed method in Neural Network (with w1

    = w2 = w3= w4= w5) on simu- lated (top) and real data (bottom).

    B.Assessment With SPM

    Statistical parametric mapping technique created for examining differences in brain activity recorded during functional neural imaging experiments using neural imaging technologies such as MRI or PET.

    Statistical parametric mapping refers to the construction of spatially extended statistical processes to test hypotheses about regionally specific effects. Statistical parametric maps (SPMs) areimage processes with voxel values that have, under the null hypothesis, a known distributional approximation (usually Gaussian). The power of statistical parametric mapping is largely due to the simplicity of the underlying idea: Namely one proceeds by analyzing each voxel using any (univariate) statistical parametric test..

    C. Artificial Neural Networks

    The network is composed of a large number of highly interconnected processing elements(neurons) working in parallel to solve a specific problem. Neural networks learn by example. There are tasks are more suited to an algorithmic approach like arithmetic operations and tasks that are more suited to neural networks. Even more, a large number of tasks, require systems that use a combination of the two approaches (normally a conventional computer). Statistical parametric mapping or SPM is a statistical technique created by Karl Freestone for examining differences in brain activity recorded during functional neuroimaging experiments using neuroimaging technologies such as MRI or PET.

  3. CONCLUSION

The constructed statistical model of normal brain images has been applied to the segmentation of brain pathologies. The assessment of the method two-group analysis performed with SPM and illustrated the potential of semisupervised learning in such applications.Finaly easy to identify the Brain Tumor is normal and abnormal. In such cases abnormal means to mentioned initial, middle and final stages of the brain tumor.

REFERENCES

  1. ] S. Shen, A. J. Szameitat, and A. Sterr, Detection of infarct lesions from single MRI modality using inconsistency between voxel intensity and spatial locationA 3-D automatic approach, IEEE Trans. Inf. Technol. Biomed., vol. 12, no. 4, pp. 532540, Jul. 2008.

  2. S. Samar, S. Boyd, and D. Gorinevsky, Distributed estimation via dual decomposition, in Proc. Eur. Control Conf., 2007, pp. 1511 1516.

  3. E. I. Zacharaki, S. Kanterakis, R. N. Bryan, and C. Davatzikos, Mea- suring brain lesion progression with a supervised tissue classification sys- tem, in Medical Image Computing and Computer-Assisted Intervention (Lecture Notes Computer Science Series), vol. 11. Berlin, Germany: Springer, 2008, pp. 620627.

  4. O. Colliot, T. Mansi, N. Bernasconi, V. Naessens, D. Klironomos, and A. Bernasconi, Segmentation of focal cortical dysplasia lesions using a feature-based level set, in Medical Image Computing and Computer- Assisted (Lecture Notes Computer Science), vol. 3749. Berlin, Ger- many: Springer, 2005, pp. 375 382.

  5. J. D. Williamson, M. E. Miller, R. N. Bryan, R. M. Lazar, L. H. Coker, J. Johnson, T. Cukierman, K. R. Horowitz, A. Murray, and

    L. J. Launer, The action to control cardiovascular risk in diabetes memory in diabetes study (ACCORD-MIND): Rationale, design, and methods, Amer. J. Cardiol., vol. 99, no. 12, pp. S112S122, 2007

  6. M. Roua¨nia, M. S. Medjram, and N. Doghmane, Brain MRI

    segmen- tation and lesions detection by EM algorithm, World Acad. Sci., Eng. Technol., vol. 24, pp. 139142, 2006.

  7. P. Aljabar, R. A. Heckemann, A. Hammers, J. V. Hajnal, and D. Rueckert, Multi-atlas based segmentation of brain images: Atlas selection and its effect on accuracy, Neuroimage, vol. 46, no. 3, pp. 726738, 2009.

  8. ] R. Verma, E. I. Zacharaki, Y. Ou, H. Cai, S. Chawla, R. Wolf, S.-K. Lee, E. R. Melhem, and C. Davatzikos, Multi-parametric tissue characteriza- tion of brain neoplasms and their recurrence using pattern classification of MR images, Acad. Radiol., vol. 15, pp. 966977, 2008.

  9. E. A. Stamatakis and L.K. Tyler, Identifying lesions on structural brain imagesValidation of the method and application to neuropsychological patients, Brain Lang., vol. 94, pp. 167177, 2005.

  10. N.-X. Lian and C. Davatzikos, Groupwise morphometric analysis based on morphological appearance manifold, in Proc. IEEE Comput. Soc. Conf. Comput. Vision Pattern Recog. Workshops, 2009, pp. 133140.

Leave a Reply

Your email address will not be published. Required fields are marked *