Asian Journal of Computer Science and Technology ISSN: 2249-0701 Vol.7 No.2, 2018, pp. 107-112

© The Research Publication, www.trp.org.in

A Novel Pre-Processing Approach for the Denoising of Alzheimer Disease

Image Dataset

M. Natarajan1 and S. Sathiamoorthy

2*

1&2Assistant Professor, Division of Computer and Information Science, Annamalai University, Annamalai Nagar, Tamil Nadu, India

E-Mail: ks_sathia@yahoo.com *Corresponding Author

Abstract - Medical imaging acts an essential part in the domain

of medicinal science. In today situation image segmentation is

employed to obtain abnormal tissues from healthy tissues

naturally in medical images. Noise in an image is annoying to

us as it degrades and interrupts the quality of the image. Noise

removal is perpetually a problematic task so as edge

maintenance at the intensity of the interrupted noise in the real

image is essential. Alzheimer's disease is a neurological

dysfunction in which the destruction of brain cells creates

cognitive decline and Memory Loss. A neurodegenerative type

of dementia, the disease begins mild and becomes progressively

severe. A vital area under medical research is Brain image

analysis, effects to identify brain diseases. The preeminent

causes of Alzheimer's diseases are blood flow and low brain

activity. In this paper, a framework has proposed in the pre-

processing of filtering and noise removal in the Alzheimer

disease image dataset. This framework optimally balances the

level of protection methods according to the noise density.

Keywords: Image processing, Alzheimer Disease, Pre-

processing, Noise removal, Filtering

I. INTRODUCTION In the country, over 400 thousand people have Alzheimer's disease (AD) [1] (approximately 6% of the population). One in six countries above age 60 have diagnosed with an AD, and it is the fourth leading cause of death in the countries. An absolute diagnosis is not possible until moderate to severe cortex damage has occurred. By using image processing techniques [2], an Alzheimer region can take by utilizing a mixture of feature extraction, classification, denoising, and segmentation techniques. The proposed method has the potential of aiding in medical examination.

II. RELATED WORKS

The authors in [3], acquire an image noise filter fitting for MRI in real time (display and acquisition), which protects tiny isolated details and efficiently eliminates environment noise without introducing blur, smearing, or patch artifacts. The authors in [4], the paper introduced a three-step framework for classifying multiclass radiography images. The first step utilizes a de-noising technique based on wavelet transform (WT) and the statistical Kolmogorov Smirnov (KS) test to remove noise and insignificant features of the images. An unsupervised deep belief network

(DBN) has created for learning the unlabeled features in the second step. The combination of WT and KS test in the first step helps to improve the performance of DBNs. Discriminative feature subsets obtained in the first two steps serve as inputs into classifiers in the third step for evaluations. The authors in [5], explained the trans-radial coronary angiography (TRA) has associated with increased radiations doses. The authors hypothesized that contemporary image reduction technology would reduce radiation doses in the cardiac catheterization laboratory in a typical clinical setting. The authors in [6], the presented algorithm consists of two stages in which the first stage detects whether impulse noise has corrupted pixels and the second stage performs a filtering operation on the detected noisy pixels. Experimentally, it has found that the proposed scheme yields a better Peak Signal-to-Noise Ratio (PSNR) compared to other existing median-based impulse noise filtering schemes. The authors in [7], the authors explained an optimal adaptive global threshold selection by maximizing between-class standard deviation through histogram peak analysis to obtain a coarse segmentation. This paper investigates a new computer-aided approach to detect the abnormalities in the digital mammograms using a Dual Stage Adaptive Thresholding (DuSAT). The authors in [8], an algorithm for multiple watermarking based on discrete wavelet transforms (DWT), discrete cosine transform (DCT) and singular value decomposition (SVD) has proposed for healthcare applications. For identity authentication purpose, the proposed method uses three watermarks in the form of medical Lump image watermark, the doctor signature/identification code and the diagnostic information of the patient as the text watermarks. The authors in [9], proposed a method which uses KNN and high boost filter. The KNN classifier is used to select skin and non-skin pixel, and then a high boost filter approach is applied which can eliminate the noise and blur from the image.

107 AJCST Vol.7 No.2 July-September 2018

The authors in [10], proposed a hybrid impulse noise filter, it has implemented in two phases, in the first phase, fuzzy rules are used to detect the pixels affected by impulse noise, and in the second phase, an artificial neural network is used to remove noise from the affected pixel. The authors in [11], proposed the Medav Filter which is a combination of mean and adaptive median filter that optimally adjusts the level of mask operations according to the noise density. The median filter has good noise removal qualities, but its complexity is undesirable.

III. ALZHEIMER DISEASE Alzheimer's is a progressive disease, where dementia symptoms gradually worsen over many years [12]. In its early stages, memory loss is mild, but with late-stage Alzheimer's, individuals lose the ability to carry on a conversation and respond to their environment. Alzheimer's is the sixth leading cause of death in the United States. Those with Alzheimer's live an average of eight years after their symptoms become noticeable to others, but survival can range from four to 20 years, depending on age and other health conditions. Alzheimer's has no current cure, but treatments for symptoms are available, and research continues. Although current Alzheimer's treatments cannot stop Alzheimer's from progressing, they can temporarily slow the worsening of dementia symptoms and improve quality of life for those with Alzheimer's and their caregivers. Today, there is a worldwide effort under way to find better ways to treat the disease, delay its onset, and prevent it from developing. Alzheimer is the total brain size shrinks, and the tissue has progressively fewer nerve cells and connections. As a result, this is cause to in lower thinking & occasional problems of remembering certain things. The Alzheimer's Association says in its early onset information that doctors do not expect to find Alzheimer’s disease in younger people. For the younger age groups, doctors will look for other dementia causes first [1][13]. Visualization of brain structure and function from the level of individual molecules to the whole brain [14]. Many imaging methods are noninvasive and allow dynamic processes to be monitor over time. Imaging is enabling researchers to identify neural networks involved in cognitive processes; understand disease pathways; recognize and diagnose diseases early, when they have most effectively treated; and determine how therapies work.

IV. PROPOSED PREPROCESSING FRAMEWORK

FOR DENOISING OF ALZHEIMER DISEASE There are various methodologies available for denoising the medical imaging by using Image processing techniques. The proposed pre-processing framework has used to denoising the AD image dataset for enhancing the accuracy of AD risk severity prediction. In this framework, to remove the noise in the image, the various techniques are utilized. Figure 1

represents the proposed pre-processing framework for denoising of AD image dataset.

Fig. 1 Proposed Pre-Processing Framework for Denoising of Alzheimer

Disease Pseudo-code for the Proposed Pre-processing framework for Denoising of Alzheimer Disease.

Step 1: Image is input to the proposed pre-processing framework. An image is resizing in 256 × 256 dimensions then converted into a gray scale image using the size method. Step 2: Histogram Equalization is applied to improve the image quality. Step 3: Image has transformed into binary and thresholding has applied. The pixels values have labeled as zero and other is one utilizing threshold.

G (x,y) = 1 f(x,y) T 0 f(x,y)

Step 4: It is beneficial for focusing regions and also loads gaps in an image. Open Morphology has employed in the proposed pre-processing framework which assists to reduce small holes that create noise. Step 5: Wavelet transform approach has utilized to exclude the noise in the AD image set.

V. RESULT AND DISCUSSION The Alzheimer Image dataset from different sources like OASIS, Kaggle. The input images have named as ADImage1, ADImage2 and ADImage3 considered for pre-processing by the proposed framework. Here in this work, Salt and Pepper noise has considered for evaluation.

A. Performance Metrics

The widely used quantitative measure peak signal-to-noise ratio (PSNR), root mean squared error (RMSE), and Structural Similarity Index Measure (SSIM).

108AJCST Vol.7 No.2 July-September 2018

M. Natarajan and S. Sathiamoorthy

1. Root Mean Squared Error (RMSE) RMSE describes the aggregate squared error among the original and restored image. Lower the value of MSE happens in less error. ( ( ) ̂( ))

√

∑ ( ( ) ̂( ))

2. Peak Signal to Noise Ratio (PSNR) PSNR, in decibels, is applied as a quality measurement between ( ), and ̂( ). Higher the PSNR effects in enhanced quality of the image. The PSNR has calculated by applying the equation

( ( ) ̂( )) (

)

Where N denotes the number of grey levels, m and n express the dimension of an original image.

B. Experimental Result and Discussion Table 1-table 6 depicts the performance analysis of the proposed framework with two much-known filters called Median Filter and Mean filter for noise removal in AD image dataset at different salt and pepper noise level from 5% to 30%. Only 3 AD images have considered for this experiment. Root Mean Squared Error value must be low, and Peak Signal to Noise Ratio value must be high. From the tables obtained, it is clear that the proposed pre-processing framework gives the better result than the using median and mean filter as the noise removal methods in AD image dataset.

TABLE II PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 5%

Image Number Performance Metrics Input Image at 5% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 9.24 9.42 8.68

PSNR 28.81 28.65 29.36

ADImage2 RMSE 10.68 10.7 10.32

PSNR 27.55 27.53 27.85

ADImage3 RMSE 11.22 10.88 10.78

PSNR 27.12 27.39 28.48

TABLE II PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 10%

Image Number Performance Metrics Input Image at 10% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 16.16 18.63 16.28

PSNR 23.84 22.73 24.95

ADImage2 RMSE 19.16 20.65 19.21

PSNR 22.48 21.83 22.46

ADImage3 RMSE 20.53 21.34 20.42

PSNR 21.77 21.55 22.00

TABLE III PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 15%

Image Number Performance Metrics Input Image at 15% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 25.10 28.59 25.19

PSNR 20.03 19.01 20.10

ADImage2 RMSE 28.4 31.07 28.74

PSNR 19.06 18.28 18.96

ADImage3 RMSE 30.69 32.41 30.80

PSNR 18.38 17.90 18.45

109 AJCST Vol.7 No.2 July-September 2018

A Novel Pre-Processing Approach for the Denoising of Alzheimer Disease Image Dataset

TABLE IV PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 20%

Image Number Performance Metrics Input Image at 20% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 34.51 39.35 35.29

PSNR 17.07 16.23 17.18

ADImage2 RMSE 38.16 48.06 38.81

PSNR 16.49 15.65 15.35

ADImage3 RMSE 43.08 44.05 41.54

PSNR 15.44 15.25 15.76

TABLE V PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 25%

Image Number Performance Metrics Input Image at 25% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 46.76 50.25 45.60

PSNR 14.72 14.11 14.95

ADImage2 RMSE 50.39 53.14 49.16

PSNR 14.08 13.62 14.39

ADImage3 RMSE 53.78 55.47 52.06

PSNR 13.52 13.25 13.80

TABLE VI PERFORMANCE ANALYSIS OF THE PROPOSED FRAMEWORK WITH OTHER TWO FILTERS FOR DENOISING OF AD IMAGE AT NOISE LEVEL OF 30%

Image Number Performance Metrics Input Image at 30% Noise Level

Median Filter Mean Filter Proposed Framework

ADImage1 RMSE 56.72 60.75 55.40

PSNR 13.05 12.46 13.26

ADImage2 RMSE 60.95 64.77 60.08

PSNR 12.43 11.9 12.56

ADImage3 RMSE 64.65 67.39 63.28

PSNR 11.91 11.56 12.11 Figure 2, 3 and 4 depicts the graphical representation of ADImage1, ADImage2, and ADImage3 performance analysis of the Median Filter, Mean Filter and proposed Pre-processing framework for Denoising of AD image dataset at various noise levels (5%, 10%, 15%, 20%, 25% and 30%) against RMSE value

Fig. 2 Graphical representation of the ADImage1 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework for Denoising of AD image dataset at various noise levels against the

RMSE value

Fig. 3 Graphical representation of the ADImage2 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework for Denoising of AD image dataset at various noise levels against the

RMSE value

110AJCST Vol.7 No.2 July-September 2018

M. Natarajan and S. Sathiamoorthy

Fig. 4 Graphical representation of the ADImage3 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework for Denoising of AD image dataset at various noise levels against the RMSE value

Figure 5, 6 and 7 depicts the graphical representation of ADImage1, ADImage2, and ADImage3 performance analysis of the Median Filter, Mean Filter and proposed Pre-processing framework for Denoising of AD image dataset at various noise levels (5%, 10%, 15%, 20%, 25% and 30%) against PSNR value.

Fig. 5 Graphical representation of the ADImage1 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework

for Denoising of AD image dataset at various noise levels against PSNR value

Fig. 6 Graphical representation of the ADImage2 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework

for Denoising of AD image dataset at various noise levels against PSNR value

Fig. 7 Graphical representation of the ADImage3 performance analysis of Median Filter, Mean Filter and Proposed Pre-processing framework

for Denoising of AD image dataset at various noise levels against PSNR value

VI. CONCLUSION

In this paper, a new pre-processing framework has proposed which is composed of different techniques for image denoising of Alzheimer Disease image dataset. The results show that the proposed framework suits well than existing filtering methods. The quantitative measurement of PSNR and RMSE shows the effectiveness of the algorithm. It indicates that the proposed framework can remove noise efficiently. The proposed framework has benefit in many quantitative techniques that rely on the quality of the data.

REFERENCES

[1] Yang, Yuan-Han, et al. "Gender’s Effects to the Early Symptoms

of Alzheimer’s Disease in 5 Asian Countries." American Journal

of Alzheimer's Disease & Other Dementias, Vol. 32, No. 4, pp. 194-199, 2017.

[2] Gulhare, Kajal Kiran, S. P. Shukla and L. K. Sharma, "Overview on segmentation and classification for the Alzheimer‟ s disease detection from brain MRI." pp. 130-132, 2017.

[3] Klosowski, Jakob and Jens Frahm, "Image denoising for real‐time MRI." Magnetic resonance in medicine, Vol. 77, No. 3, 1340-1352, 2017.

[4] Khatami, Amin, et al. "Medical image analysis using wavelet transform and deep belief networks." Expert Systems with Applications, Vol. 86, pp. 190-198, 2017.

[5] Gunja, Ateka, et al. "Image noise reduction technology reduces radiation in a radial-first cardiac catheterization laboratory." Cardiovascular Revascularization Medicine, Vol. 18, No. 3, pp.197-201, 2017.

[6] Bhadouria, Vivek Singh, et al. "A novel image impulse noise removal algorithm optimized for hardware accelerators." Journal

of Signal Processing Systems, Vol. 89, No. 2, pp. 225-242, 2017. [7] J. Anitha, J. Dinesh Peter and S. Immanuel Alex Pandian. "A dual

stage adaptive thresholding (DuSAT) for automatic mass detection in mammograms." Computer methods and programs in biomedicine, Vol. 138, pp. 93-104, 2017.

[8] Zear, Aditi, Amit Kumar Singh, and Pardeep Kumar. "A proposed secure multiple watermarking technique based on DWT, DCT and SVD for application in medicine." Multimedia Tools and

Applications, Vol. 77, No. 4, pp. 4863-4882, 2018. [9] Vinay Singh and Deepa Aswani, "Face Detection in Hybrid Color

Space Using HBF-KNN." Proceedings of International

Conference on Recent Advancement on Computer and

Communication. Springer, Singapore, 2018. [10] Khwairakpam Amitab, et al. "Impulse Noise Reduction in Digital

Images Using Fuzzy Logic and Artificial Neural Network." Proceedings of the International Conference on

111 AJCST Vol.7 No.2 July-September 2018

A Novel Pre-Processing Approach for the Denoising of Alzheimer Disease Image Dataset

Computing and Communication Systems. Springer, Singapore, 2018.

[11] Sayantan Gupta and Sukanya Roy, "Medav Filter—Filter for Removal of Image Noise with the Combination of Median and Average Filters." Recent Trends in Signal and Image Processing. Springer, Singapore, pp. 11-19, 2019.

[12] Ying Liang and Lei Wang. "Alzheimer’s disease is an important risk factor of fractures: a meta-analysis of cohort studies." Molecular neurobiology, Vol. 54, No.5, pp. 3230-3235, 2017.

[13] Deepshikha Bhardwaj, et al. "Alzheimer's disease—Current Status and Future Directions." Journal of medicinal food, Vol. 20, No.12, pp. 1141-1151, 2017.

[14] Hao He, et al. "Co-altered functional networks and brain structure in unmedicated patients with bipolar and major depressive disorders." Brain Structure and Function, Vol. 222, No. 9, pp. 4051-4064, 2017.

.

112AJCST Vol.7 No.2 July-September 2018

M. Natarajan and S. Sathiamoorthy