مشارکت سه بافت مغزی در تشخیص بیماری آلزایمر از MRI ساختاری
الموضوعات :شیما تاج الدینی 1 , حبیب اله دانیالی 2 , محمدصادق هل فروش 3 , یعقوب فاطمی 4
1 - Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran
2 - Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran
3 - Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran
4 - Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran
الکلمات المفتاحية: بیماری آلزایمر, نشانگر زیستی, طبقه بندی, استخراج ویژگی, تصویربرداری رزونانس مغناطیسی,
ملخص المقالة :
بیماری آلزایمر (AD) یک بیماری پیشرونده و برگشت ناپذیر است که به تدریج باعث می شود بیماران نتوانند کارهای روزمره خود را انجام دهند. اگرچه روش های درمانی فعلی نمی توانند بیماری را به طور کامل درمان کنند ، اما تشخیص به موقع آن می تواند علائم را کاهش داده و کیفیت زندگی بیماران را افزایش دهد. در ادبیات فعلی ، استفاده از بافت ماده خاکستری (GM) که به عنوان نشانگر زیستی مناسب شناخته می شود ، در تشخیص AD بسیار رایج است. با این حال ، به نظر می رسد دو بافت مغز دیگر معروف به مایع مغزی نخاعی (CSF) و ماده سفید (WM) اطلاعات مفیدی را درباره تغییرات مغزی بیماران نشان می دهند. هدف از مطالعه حاضر ایجاد یک سیستم اتوماتیک برای تشخیص زود هنگام بیماری آلزایمر از MRI ساختاری با در نظر گرفتن همزمان ویژگی های مناسب از تمام بافت های GM ، CSF و WM است. یک طبقه بندی SVM-RBF بر روی پایگاه داده OASIS آموزش داده شده و مورد ارزیابی قرار می گیرد تا AD از افراد سالم کنترل شود. نتایج به دست آمده نشان دهنده دقت و حساسیت بالاتر الگوریتم پیشنهادی در مقایسه با روش مشابه است
[1] Alzheimer’s disease and Dementia Alzheimer’s Association, What is Alzheimer’s disease, http://alz.org/alzheimers_disease_what_is_alzheimers.asp (accessed 25.06.16).
[2] F. Falahati, E. Westman, and A. Simmons, ‘’Multivariate data analysis and machine learning in Alzheimer's disease with a focus on structural magnetic resonance imaging,’’ Journal of Alzheimer's disease: JAD, vol. 41, pp. 685-708, 2012.
[3] M. Atiya, B. Hyman, M. Albert, and R. Killiany, ‘’Structural magnetic resonance imaging in established and prodromal Alzheimer disease: a review,’’ Alzheimer Disease & Associated Disorders, Vol. 17, pp. 177-195, 2003.
[4] S. Kazemifar, J. Drozd, N. Rajakumar, M. Borrie, and R. Bartha, "Automated algorithm to measure changes in medial temporal lobe volume in Alzheimer disease," Journal of neuroscience methods, vol. 227, pp. 35-46, 2014.
[5] R. Wolz, V. Julkunen, J. Koikkalainen, E. Niskanen, D. Zhang, D. Rueckert, H. Soininen, and J. Lötjönen, ‘’Multi-method analysis of MRI images in early diagnostics of Alzheimer's disease,’’ PloS one, vol. 6, p.e25446, 2011.
[6] S. Costafreda, I. Dinov, Z. Tu, Y. Shi, C. Liu, I. Kloszewska, P. Mecocci, H. Soininen, M. Tsolaki, B. Vellas, and L. Wahlund, ‘’Automated hippocampal shape analysis predicts the onset of dementia in mild cognitive impairment,’’ Neuroimage, Vol. 56, pp. 212-219, 2011.
[7] K. Shen, J. Fripp, F. Mériaudeau, G. Chételat, O. Salvado, and P. Bourgeat, ‘’Detecting global and local hippocampal shape changes in Alzheimer's disease using statistical shape models,’’ Neuroimage, vol. 59, pp. 2155-2166, 2012.
[8] C. Aguilar, E. Westman, J. Muehlboeck, P. Mecocci, B. Vellas, M. Tsolaki, I. Kloszewska, H. Soininen, S. Lovestone, C. Spenger, and A. Simmons, ‘’Different multivariate techniques for automated classification of MRI data in Alzheimer’s disease and mild cognitive impairment,’’ Psychiatry Research: Neuroimaging, vol. 212, pp. 89-98, 2013.
[9] O. Colliot, G. Chételat, M. Chupin, B. Desgranges, B. Magnin, H. Benali, B. Dubois, L. Garnero, F. Eustache, and S. Lehéricy, ‘’Discrimination between Alzheimer disease, Mild Cognitive Impairment, and Normal Aging by Using Automated Segmentation of the Hippocampus 1,’’ Radiology, vol. 248, pp. 194-201, 2008.
[10] T. Stoub and B. Dickerson, ‘’Parahippocampal white matter volume predicts Alzheimer's disease risk in cognitively normal old adults,’’ Neurobiology of aging, vol. 35, pp. 1855-1861, 2014.
[11] C. Wang, G. Stebbins, D. Medina, R. Shah, R. Bammer, M. Moseley, and L. deToledo-Morrell, ‘’Atrophy and dysfunction of parahippocampal white matter in mild Alzheimer’s disease,’’ Neurobiol. Aging, vol. 33, pp. 43e52, 2012.
[12] D. Salat, D. Greve, J. Pacheco, B. Quinn, K. Helmer, R. Buckner, and B. Fischl, ‘’Regional white matter volume differences in nondemented aging and Alzheimer’s disease,’’ Neuroimage, vol. 44, pp. 1247e1258, 2009.
[13] D. Salat, D. Tuch, A. Van der Kouwe, D. Greve, V. Pappu, S. Lee, N. Hevelone, A. Zaleta, J. Growdon, S. Corkin, and B. Fischl, ‘’White matter pathology isolates the hippocampal formation in Alzheimer's disease,’’ Neurobiology of aging, vol. 31, pp.244-256, 2010.
[14] S. Madsen, B. Gutman, S. Joshi, A. Toga, C. Jack, M. Weiner, and P. Thompson, ‘’Mapping ventricular expansion onto cortical gray matter in older adults,’’ Neurobiology of aging, vol. 36, pp. S32-S41, 2015.
[15] S. Nestor, R. Rupsingh, M. Borrie, M. Smith, V. Accomazzi, J. Wells, J. Fogarty, and R. Bartha, ‘’Ventricular enlargement as a possible measure of Alzheimer's disease progression validated using the Alzheimer's disease neuroimaging initiative database,’’ Brain, vol. 131, pp. 2443-2454, 2008.
[16] L. Apostolova, A. Green, S. Babakchanian, K. Hwang, Y. Chou, A. Toga, and P. Thompson, ‘’Hippocampal atrophy and ventricular enlargement in normal aging, mild cognitive impairment and Alzheimer's disease,’’ Alzheimer disease and associated disorders, vol. 26, p. 17, 2012.
[17] L. Clerx, I. A. van Rossum, L. Burns, D. L .Knol, P. Scheltens, F. Verhey, P. Aalten, P. Lapuerta,L. van de Pol, R. van Schijndel, and R. de Jong, ‘’Measurements of medial temporal lobe atrophy for prediction of Alzheimer's disease in subjects with mild cognitive impairment,’’ Neurobiology of aging, vol. 34, pp. 2003-2013, 2013.
[18] D. Marcus, T. Wang, J. Parker, J. Csernansky, J. Morris, and R. Buckner, ‘’Open Access Series of Imaging Studies (OASIS): cross-sectional MRI data in young, middle aged, nondemented, and demented older adults,’’ Journal of cognitive neuroscience, Vol. 19, pp.1498-1507, 2007.
[19] G A. Papakostas, A. Savio, M. Graña, V. G. Kaburlasos, ‘’A lattice computing approach to Alzheimer’s disease computer assisted diagnosis based on MRI data,’’ Neurocomputing, vol. 150, pp. 37-42, 2015.
[20] K. Juottonen, M. Laakso, K. Partanen, and H. Soininen, ‘’Comparative MR analysis of the entorhinal cortex and hippocampus in diagnosing Alzheimer disease,’’ American Journal of Neuroradiology, Vol. 20, pp. 139-144, 1999.
[21] N. Tzourio-Mazoyer, B. Landeau, D. Papathanassiou, F. Crivello, O. Etard, N. Delcroix, B. Mazoyer, and M. Joliot, ‘’Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain,’’ Neuroimage, Vol. 15, pp. 273-289, 2002.
[22] E. Atashpaz-Gargari, and C. Lucas, ‘’Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition,’’ Evolutionary Computation, IEEE Congress on, pp. 4661-4667, 2007.
Iranian Journal of Information Technology & Communication | No.33-34, Vol.9, Fall & Winter 2018 PP: 9-14 |
|
The Participation of Three Brain Tissues in Alzheimer’s disease Diagnosis from Structural MRI
* Shima Tajeddini * Habibollah Danyali *Mohammad Sadegh Helfroush * Yaghoub Fatemi
* Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran
Abstract
Alzheimer’s disease (AD) is a progressive and irreversible disease which gradually makes patients unable to do their daily routines. Although the present treatments can not cure the disease completely, its early detection can reduce symptoms and enhance the patients’ life quality. In the current literature, using the grey matter (GM) tissue which is known as an appropriate biomarker is highly common in AD diagnosis. However, two other brain tissues known as cerebrospinal fluid (CSF) and white matter (WM) seem to reveal beneficial information about the patients’ brain changes. The aim of the present study is to develop an automatic system for the early diagnosis of Alzheimer’s disease from structural MRI by simultaneously considering suitable features of all GM, CSF and WM tissues. A SVM-RBF classifier is trained and evaluated on the OASIS database to separate AD from healthy control (HC) subjects. The obtained results represent higher accuracy and sensitivity of the proposed algorithm in comparison with similar method.
Keywords: Alzheimer’s disease; Biomarker; Classification; Feature extraction; Magnetic resonance imaging
I. Introduction
Alzheimer is a neurological disease which attacks the brain neurons and causes the brain neurodegeneration. Most of the time, its early symptom is the short term memory loss. However, Alzheimer’s disease gradually causes some sever symptoms like changes in mood, speech, though and reasoning power. Human brain is made of nearly 100 billion neurons and scientists believe that any change in these neurons or their connections causes the Alzheimer’s disease. Indeed, Alzheimer is the sixth death factor in the United States. It is apt to mention that increasing the age is the major risk factor in afflicting Alzheimer and may engage most of people who are above 65. However, people may be afflicted with AD during their forties or fifties [1]. Finding brain suitable biomarkers helps experts to AD detection by means of the computer aided diagnosis (CAD) systems. Besides, due to their high resolution and contrast, MRI images are considered as suitable means of finding these biomarkers [2].
|
In the present paper, an automatic system which is designed based on the participation of three brain tissues (GM, WM and CSF) is proposed to simultaneously take the advantages of these tissues in AD diagnosis.
The rest of this paper is organized as follows: in Section 2, the used database is explained. The proposed algorithm is presented in Section 3. Section 4 encompasses experimental results and discussion and finally, Section 5 concludes the paper.
2. Materials
T1-weighted MRI images needed for the algorithm evaluation are derived from the standard OASIS database [18]. This database includes cross-sectional collection of 416 individuals. Chosen individuals are divided into 49 AD and 49 HC subjects. To have a fair comparison, the number and the health condition of these individuals are selected the same with those who participated in [19]. In Table 1, these individuals’ characteristics are listed.
Healthy control (HC) | Alzheimer’s disease (AD) | _ |
49 | 49 | number |
77.77 (65-94) | 78.06 (66-96) | age |
0 | 31/17/1 | CDR (0.5/1/2) |
28.9 (26-30) | 24 (15-30) | MMSE |
Table 1: The used data derived from the OASIS database. The best CDR=0, and the worst CDR=2. The worst MMSE=15 and the best MMSE=30.
3. Methods
The blocked diagram of the proposed system to separate AD from HC subjects is shown in Figure 1. This system includes three steps: pre-processing, feature extraction and classification.
Figure 1: Block diagram of the proposed system to separate AD patients from HC subjects.
3.1. Pre-processing step
All pre-processing steps are done by the SPM software version 12 and the VBM software version 8. Before segmenting brain tissues, first all images are normalized. Each image is segmented into its three tissues known as GM, WM, and CSF. Furthermore, it must be noted that the inter-subject alignment is done by DARTEL. All GM images are modelled that help to preserve the total GM volume. In Figure 2, three segmented tissues of the brain are represented.
Figure 2: Segmenting of an AD patient into its three tissues: respectively from the right side there are GM, WM, and CSF.
3.2. Feature extraction step
As mentioned, hippocampus is prone to be afflicted with atrophy at the early stages of the Alzheimer’s disease; thus, this structure feature extraction is highly helpful and beneficial. Although the manual segmentation of the hippocampus is the most precise method [20], it is highly time consuming. Therefore, in the present article an automatic atlas-based method is used. Additionally, the Automatic Anatomic Labelling (AAL) atlas is utilized for segmenting the hippocampus [21].
Lateral ventricle is clearly visible from the brain CSF tissue. Noticeably, the ventricle volume is bigger in AD patients in relation to HC subjects. This incident is occurred as a result of shrinking of two other tissues of GM and WM. To segment this structure, the ALVIN toolbox is used. This algorithm is completely automatic, run under SPM software and can segment the lateral ventricle from MRI images. Figure 3 shows the lateral ventricle enlargement in AD and HC subjects.
Figure 3: Sample of brain lateral ventricle. a) HC subject, b) mild AD patient.
Totally, the applied features are as follows:
· Features extracted from the GM include: a) the hippocampus + the ventricle volume b) the mean and the standard deviation of the GM
· Features extracted from the WM tissue which include the WM mean and standard deviation
· Features extracted from the CSF which include the lateral ventricle enlargement
Using these features, efficient biomarkers are extracted from the most regions of the brain and the participation process of three brain tissues is fulfilled. This task can improve the performance of the proposed system.
3.3. Classification step
In the proposed system a support vector machine (SVM) classifier is employed. SVM occupies an elegant role in the literature and could accurately separate AD from HC subjects. The considerable fact regarding the SVM classifier operation is its structural risk minimization strategy which is handled to arrive at a hyper plane and consequently maximize the distance between training classes. The produced feature vector is given to this classifier as an input. The RBF kernel is used in the present study.
4. Experimental results and discussions
Cross validation (CV) is known as a statistical method to evaluate and compare classifiers. In CV, a part of database is chosen to train and another part is selected to test the classifier. Elaborately, k-fold CV is a common method to evaluate the performance of the classifier. This task reduces the evaluation variance. In the present study, samples are divided into 10 parts and then they are trained and evaluated for 10 times. Finally, their output results are averaged. It should be mentioned that the number of AD patients and HC subjects are considered equal. To evaluate the algorithm performance, some criteria have been considered like accuracy (ACC) and sensitivity (SEN).
Cross validation (CV) is known as a statistical method to evaluate and compare classifiers. In CV, a part of database is chosen to train and another part is selected to test the classifier. Elaborately, k-fold CV is a common method to evaluate the performance of the classifier. This task reduces the evaluation variance. In the present study, samples are divided into 10 parts and then they are trained and evaluated for 10 times. Finally, their output results are averaged. It should be mentioned that the number of AD patients and HC subjects are considered equal. To evaluate the algorithm performance, some criteria have been considered like accuracy (ACC) and sensitivity (SEN).
(1)
(2)
Table 2 shows the results of investigating the effect of each brain tissue in separating AD from HC subjects. For this purpose, features related to each tissue are separately given to the SVM-RBF classifier to determine the effects of each tissue in classification. As expected, the GM tissue has the greatest participation in AD diagnosis. However, two other tissues also have effective roles in separating patients; therefore, three tissues’ participation could improve the ACC and SEN of the proposed system.
SEN (%) | ACC (%) | AD/HC | Tissues |
75.6 | 81.7 | 49/49 | GM |
71.5 | 62.3 | 49/49 | WM |
51.1 | 63.3 | 49/49 | CSF |
85.8 | 86.8 | 49/49 | All tissues |
Table 2: Proposed system performance on GM, WM, and CSF tissues
In Table 3 a comparison result of the proposed algorithm performance with similar method which is presented in [19] is shown. To have a fair comparison, all features of the used database are same with this work. Papakostas et al. [19] have used a new k-NN classifier named Lattice Computing (LC) to reduce the dimension of features and to classify AD and HC subjects. As the results show, the proposed system overcomes the method in [19] in both ACC and SEN.
SEN (%) | ACC (%) | AD/HC | Database | Methods |
78 | 85 | 49/49 | OASIS | Papakostas et al. [19] |
85.8 | 86.8 | 49/49 | OASIS | Proposed method |
Table 3: The comparison of the classification results in separating AD from HC subjects on MRI images.
5. Conclusions
Early diagnosis of Alzheimer’s disease can increase the life quality of AD patients. In the present article, an automatic system was proposed to diagnose the Alzheimer’s disease. In this system, the participation of three brain tissues known as GM, WM, and CSF was considered since all these three tissues are prone to be affected by changes in Alzheimer’s disease; therefore, each tissue can play a role of a good biomarker. The proposed method is tested on 98 MRI images derived from the OASIS database. Results have shown that the proposed system has an accepted performance in classifying AD from HC subjects and can successfully reach at 86.8% accuracy.
References
[1] Alzheimer’s disease and Dementia Alzheimer’s Association, What is Alzheimer’s disease, http://alz.org/alzheimers_disease_what_is_alzheimers.asp (accessed 25.06.16).
[2] F. Falahati, E. Westman, and A. Simmons, ‘’Multivariate data analysis and machine learning in Alzheimer's disease with a focus on structural magnetic resonance imaging,’’ Journal of Alzheimer's disease: JAD, vol. 41, pp. 685-708, 2012.
[3] M. Atiya, B. Hyman, M. Albert, and R. Killiany, ‘’Structural magnetic resonance imaging in established and prodromal Alzheimer disease: a review,’’ Alzheimer Disease & Associated Disorders, Vol. 17, pp. 177-195, 2003.
[4] S. Kazemifar, J. Drozd, N. Rajakumar, M. Borrie, and R. Bartha, "Automated algorithm to measure changes in medial temporal lobe volume in Alzheimer disease," Journal of neuroscience methods, vol. 227, pp. 35-46, 2014.
[5] R. Wolz, V. Julkunen, J. Koikkalainen, E. Niskanen, D. Zhang, D. Rueckert, H. Soininen, and J. Lötjönen, ‘’Multi-method analysis of MRI images in early diagnostics of Alzheimer's disease,’’ PloS one, vol. 6, p.e25446, 2011.
[6] S. Costafreda, I. Dinov, Z. Tu, Y. Shi, C. Liu, I. Kloszewska, P. Mecocci, H. Soininen, M. Tsolaki, B. Vellas, and L. Wahlund, ‘’Automated hippocampal shape analysis predicts the onset of dementia in mild cognitive impairment,’’ Neuroimage, Vol. 56, pp. 212-219, 2011.
[7] K. Shen, J. Fripp, F. Mériaudeau, G. Chételat, O. Salvado, and P. Bourgeat, ‘’Detecting global and local hippocampal shape changes in Alzheimer's disease using statistical shape models,’’ Neuroimage, vol. 59, pp. 2155-2166, 2012.
[8] C. Aguilar, E. Westman, J. Muehlboeck, P. Mecocci, B. Vellas, M. Tsolaki, I. Kloszewska, H. Soininen, S. Lovestone, C. Spenger, and A. Simmons, ‘’Different multivariate techniques for automated classification of MRI data in Alzheimer’s disease and mild cognitive impairment,’’ Psychiatry Research: Neuroimaging, vol. 212, pp. 89-98, 2013.
[9] O. Colliot, G. Chételat, M. Chupin, B. Desgranges, B. Magnin, H. Benali, B. Dubois, L. Garnero, F. Eustache, and S. Lehéricy, ‘’Discrimination between Alzheimer disease, Mild Cognitive Impairment, and Normal Aging by Using Automated Segmentation of the Hippocampus 1,’’ Radiology, vol. 248, pp. 194-201, 2008.
[10] T. Stoub and B. Dickerson, ‘’Parahippocampal white matter volume predicts Alzheimer's disease risk in cognitively normal old adults,’’ Neurobiology of aging, vol. 35, pp. 1855-1861, 2014.
[11] C. Wang, G. Stebbins, D. Medina, R. Shah, R. Bammer, M. Moseley, and L. deToledo-Morrell, ‘’Atrophy and dysfunction of parahippocampal white matter in mild Alzheimer’s disease,’’ Neurobiol. Aging, vol. 33, pp. 43e52, 2012.
[12] D. Salat, D. Greve, J. Pacheco, B. Quinn, K. Helmer, R. Buckner, and B. Fischl, ‘’Regional white matter volume differences in nondemented aging and Alzheimer’s disease,’’ Neuroimage, vol. 44, pp. 1247e1258, 2009.
[13] D. Salat, D. Tuch, A. Van der Kouwe, D. Greve, V. Pappu, S. Lee, N. Hevelone, A. Zaleta, J. Growdon, S. Corkin, and B. Fischl, ‘’White matter pathology isolates the hippocampal formation in Alzheimer's disease,’’ Neurobiology of aging, vol. 31, pp.244-256, 2010.
[14] S. Madsen, B. Gutman, S. Joshi, A. Toga, C. Jack, M. Weiner, and P. Thompson, ‘’Mapping ventricular expansion onto cortical gray matter in older adults,’’ Neurobiology of aging, vol. 36, pp. S32-S41, 2015.
[15] S. Nestor, R. Rupsingh, M. Borrie, M. Smith, V. Accomazzi, J. Wells, J. Fogarty, and R. Bartha, ‘’Ventricular enlargement as a possible measure of Alzheimer's disease progression validated using the Alzheimer's disease neuroimaging initiative database,’’ Brain, vol. 131, pp. 2443-2454, 2008.
[16] L. Apostolova, A. Green, S. Babakchanian, K. Hwang, Y. Chou, A. Toga, and P. Thompson, ‘’Hippocampal atrophy and ventricular enlargement in normal aging, mild cognitive impairment and Alzheimer's disease,’’ Alzheimer disease and associated disorders, vol. 26, p. 17, 2012.
[17] L. Clerx, I. A. van Rossum, L. Burns, D. L .Knol, P. Scheltens, F. Verhey, P. Aalten, P. Lapuerta,L. van de Pol, R. van Schijndel, and R. de Jong, ‘’Measurements of medial temporal lobe atrophy for prediction of Alzheimer's disease in subjects with mild cognitive impairment,’’ Neurobiology of aging, vol. 34, pp. 2003-2013, 2013.
[18] D. Marcus, T. Wang, J. Parker, J. Csernansky, J. Morris, and R. Buckner, ‘’Open Access Series of Imaging Studies (OASIS): cross-sectional MRI data in young, middle aged, nondemented, and demented older adults,’’ Journal of cognitive neuroscience, Vol. 19, pp.1498-1507, 2007.
[19] G A. Papakostas, A. Savio, M. Graña, V. G. Kaburlasos, ‘’A lattice computing approach to Alzheimer’s disease computer assisted diagnosis based on MRI data,’’ Neurocomputing, vol. 150, pp. 37-42, 2015.
[20] K. Juottonen, M. Laakso, K. Partanen, and H. Soininen, ‘’Comparative MR analysis of the entorhinal cortex and hippocampus in diagnosing Alzheimer disease,’’ American Journal of Neuroradiology, Vol. 20, pp. 139-144, 1999.
[21] N. Tzourio-Mazoyer, B. Landeau, D. Papathanassiou, F. Crivello, O. Etard, N. Delcroix, B. Mazoyer, and M. Joliot, ‘’Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain,’’ Neuroimage, Vol. 15, pp. 273-289, 2002.
[22] E. Atashpaz-Gargari, and C. Lucas, ‘’Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition,’’ Evolutionary Computation, IEEE Congress on, pp. 4661-4667, 2007.
