Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак
In this paper, the problem of choosing the right feature for diagnosing Dementia is discussed. Several features that could affect dementia were reviewed and their importance was evaluated. Random forest algorithm and SVM for the dementia diagnosis have been developed and investigated. Experiments we...
Gespeichert in:
| Datum: | 2019 |
|---|---|
| Hauptverfasser: | , |
| Format: | Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2019
|
| Schlagworte: | |
| Online Zugang: | https://journal.iasa.kpi.ua/article/view/175552 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Institution
System research and information technologies| _version_ | 1866302553223856128 |
|---|---|
| author | Naderan, Maryam Zaychenko, Yuriy P. |
| author_facet | Naderan, Maryam Zaychenko, Yuriy P. |
| author_sort | Naderan, Maryam |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2019-08-27T22:12:50Z |
| description | In this paper, the problem of choosing the right feature for diagnosing Dementia is discussed. Several features that could affect dementia were reviewed and their importance was evaluated. Random forest algorithm and SVM for the dementia diagnosis have been developed and investigated. Experiments were conducted on the open-source database and compared with the related works’ results. The purpose of the paper is to improve the accuracy of diagnosis of dementia using the reduction of features' dimension. This article is devoted to analysis of the main distinguishing features of Alzheimer`s dementia, applicable methods and treatment of Alzheimer's dementia on early stage that could help to avoid negative consequences connected with progress of the disease. The purpose of the paper is to improve the accuracy of diagnosis of dementia. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2019.2.03 |
| first_indexed | 2025-07-17T10:26:04Z |
| format | Article |
| fulltext |
Maryam Naderan, Yuriy Zaychenko, 2019
Системні дослідження та інформаційні технології, 2019, № 2 25
UDC 004.855.5
DOI: 10.20535/SRIT.2308-8893.2019.2.03
METHODS FOR IMPROVING ACCURACY OF THE DEMENTIA
DIAGNOSIS USING FEATURE DIMENSION REDUCTION
MARYAM NADERAN, YURIY ZAYCHENKO
Abstract. In this paper, the problem of choosing the right feature for diagnosing
Dementia is discussed. Several features that could affect dementia were reviewed
and their importance was evaluated. Random forest algorithm and SVM for the de-
mentia diagnosis have been developed and investigated. Experiments were con-
ducted on the open-source database and compared with the related works’ results.
The purpose of the paper is to improve the accuracy of diagnosis of dementia using
the reduction of features' dimension. This article is devoted to analysis of the main
distinguishing features of Alzheimer`s dementia, applicable methods and treatment
of Alzheimer`s dementia on early stage that could help to avoid negative conse-
quences connected with progress of the disease. The purpose of the paper is to im-
prove the accuracy of diagnosis of dementia.
Keywords: diagnosis Alzheimer’s disease, ensemble learning methods, classifica-
tion, Convolutional Neural Network.
INTRODUCTION
This article is devoted to analysis of the main distinguishing features of Alz-
heimer’s dementia and applicable methods, main elements of successful diagnos-
tics and treatment of Alzheimer’s dementia on early stage, and review of the Alz-
heimer’s association experience, that could help to avoid negative consequences
connected with progress of the disease. In addition, it’s necessary to find impor-
tant features which play the big role in Alzheimer’s disease diagnostics using by
Random Forest Algorithms and Support Vector Machine (SVM) Algorithm. Dis-
tinguishing features of Alzheimer’s dementia could be found using physical &
neurological exam, mental status tests, computerized tests, screening, genetic tests
and etc. For this research, Kaggle’s data about patients Nondemented and De-
mented was used.
General features of Alzheimer’s disease were described by foreign scientists
S. Sarraf, G. Tofighi and J. Anderson in publication devoted to classification of
the considered features using convolutional neural networks, that contain results
necessary for distinguishing classification criteria and significant features taken
into consideration for the patients with the mentioned disease. Authors stated, that
“early detection and classification of Alzheimer’s disease are critical for proper
treatment and preventing brain tissue damage. Alzheimer’s disease has a certain
progressive pattern of brain tissue damage. It shrinks the hippocampus and cere-
bral cortex of the brain and enlarges the ventricles [6]. The considered general
characteristics could be considered as common for many other mental diseases,
but the underlined damage to the brain tissue and peculiarities of negative impact
on the ventricles provide the grounds for differentiation at site.
Maryam Naderan, Yuriy Zaychenko
ISSN 1681–6048 System Research & Information Technologies, 2019, № 2 26
Approaches as to the stages of Alzheimer’s disease. Modern Ukrainian and
foreign scholars have different views as to the issue of the main stages of Alz-
heimer’s dementia, for example, in Ukrainian practice we found the approach
providing differentiation into four main stages: prior dementia, early dementia,
moderated dementia and final dementia, according to the classification provided
by Ukrainian Clinics of modern neurology “Aximed”.
RELATED WORK
Special attention should be paid to modern steps in differentiation of special types
of Alzheimer’s dementia revealed by L.V. Zhdaneeva in PhD thesis devoted to
the issues of “Movement disorders of Alzheimer’s disease” [1]. Author substanti-
ated important statistical data, regarding internal differentiation of special fea-
tures, common to patients with Alzheimer’s dementia: “87,7 % of patients with
Alzheimer’s disease had movement disorders, including: 18,9 % — a kinetic-rigid
syndrome; 27,8 % — hyperkinesias syndrome without rigidity; 8,8 % — tremor;
12,2 % — walking and postural stability disorder syndrome; 76,6 — pseudo bul-
bar syndrome; 45,6 % — stereotypes; 12,2 % — pyramidal syndrome; 6,7 % —
epileptic seizures” [1, p. 23]. These data show strong correlation between general
and special features, disorders common not only to patients with Alzheimer’s de-
mentia, that could be used for further scientific research and practical clinical ac-
tivities, as the obtained data represent a wide range of patients with Alzheimer’s
disease and provide solutions for facilitation of differentiation of patient’s status.
Modern research results in the field show, that the leading Harvard team was
“the first to try to combine fMRI scans and deep learning into a program that
could predict an MCI patient’s chance of developing Alzheimer’s disease. The
fMRI scans used in their analysis were taken when patients were at rest. As with
any fMRI scans, they reveal where electrical signals are flashing in the brain and
how these areas relate to one another”. But one of the latest publication in IEEE
underlined, that there is no such a method found yet to provide 90 and more per-
cent in recognition of Alzheimer’s dementia. Such a conclusion was made by
Dinggang Shen, famous scientist from University of North Carolina: “Nobody in
the field can get from 80 to 90 percent,” he says. “That's impossible, just based on
this method” [5].
Negative impact on patient’s memory can also be considered as a general
feature for differentiation of the status of Alzheimer’s patients. Progress of the
disease makes it impossible to realize mental function as previously, during active
adult or youth periods, even daily activities cannot be fulfilled without toil, pa-
tients thus need special assistance and support, understanding of their needs and
problems. Scientists consider Alzheimer’s disease as a severe disorder, that has
neurological nature and affects human brains in a special way described above.
The mentioned approach differs from the one provided by J. Islam and Y.
Zhang in the above-mentioned publication, providing classification into three
main stages of Alzheimer’s disease: very mild, mild and moderate [4]. The dis-
tinction between the considered approaches lies in the attitude to the prior stage of
the disease that often shows very close features to other common mental, psycho-
logical, general diseases, in this respect it is important to discover modern meth-
ods of early diagnostics and treatment of Alzheimer’s dementia, as it could pro-
vide the best results for evaluation of patient’s health.
Methods for improving accuracy of the dementia diagnosis using feature dimension reduction
Системні дослідження та інформаційні технології, 2019, № 2 27
COMPARATIVE ANALYSIS
In [2] was carried out comparative analysis about methods which are using in
medical diagnosis. It was emphasized that Convolutional Neural Network’s diag-
nosis ability is greater than endoscopists in general. In [3], authors have compared
CNN and endoscopists. The results obtained show that the sensitivity, specificity,
accuracy, and diagnostic time were 88,9%, 87,4%, 87,7%, and 194 s, respec-
tively, for Convolutional Neural Network. The same indicators obtained for the
23 endoscopists were 79,0%, 83,2% and 82,4% respectively.
Among foreign scholars and researchers we should address to modern scien-
tific articles of Jyoti Islam, Yanqing Zhang “An Ensemble of Deep Convolutional
Neural Networks for Alzheimer's Disease Detection and Classification” authors
substantiated “a novel Alzheimer's Disease detection and classification model
using brain MRI data analysis, developed an ensemble of deep convolutional neu-
ral networks and demonstrated superior performance on the Open Access Series
of Imaging Studies dataset” [4]. This aspect has special importance for domestic
research and clinical practice as provides new knowledge and methods for devel-
opment of early diagnostics of Alzheimer’s dementia.
Modern computerized systems provide experts with detailed information on
the nature of dementia, it’s stage, or precisely define, a contrario, Nondemented
status. For the past ten years significant development was achieved in using MRI,
but modern foreign researchers insist on automated brain MRI results, providing
at site segmentation and classification, comparison with other groups of patients,
i.e. control groups collected in computerized memory. This could be an example
of conclusions and recommendations of the 31-st Conference held in 2017 on
Neural Information Processing Systems in USA, where it was proposed to use
“handcrafted feature generation and extraction from the MRI data, improvement
of machine learning models such as Support Vector Machine, Logistic regression
model etc” [4, p. 2].
RESULTS
In related works, authors hadn’t mentioned which attributes are more informative
for classification. In this paper, three more important features for dementia classi-
fication are detected.
Table 1 illustrates some of the data which were used in this paper [7]. The
origin of the data set consists several visits of each patients. We consider only
those patients who visited for the first time, since during second and third visit,
patients could use medicine to prevent degradation of the dementia.
T a b l e 1 . Attributes with their definition
Attrib-
utes
Subject ID MRI ID Group Visit MR Delay M/F Hand
Definition
Subject
identification
MRI
identification
Class
Visit
followup
MR
contrast
Gender
Dominant
hand
Attributes EDUC SES MMSE CDR eTIV nWBV ASF
Definition Education
level
SES
cognition
test
MMSE
cognition
test
CDR
cogni-
tion test
Estimated
intracranial
volume
Standardized
brain
volume
Atlas
factor
scaling
Maryam Naderan, Yuriy Zaychenko
ISSN 1681–6048 System Research & Information Technologies, 2019, № 2 28
T a b l e 2 . Data of some patients during medical examination
Group Visit
MR
Delay
M/F Hand Age EDUC SES MMSE CDR eTIV nWBV ASF
Nondemented 1 0 M R 87 14 2 27 0 1987 0,696 0,883
Demented 1 0 M R 75 12 23 0,5 1678 0,736 1,046
Nondemented 1 0 F R 88 18 3 28 0 1215 0,71 1,444
Nondemented 1 0 M R 80 12 4 28 0 1689 0,712 1,039
Demented 1 0 M R 71 16 28 0,5 1357 0,748 1,293
Nondemented 1 0 F R 93 14 2 30 0 1272 0,698 1,380
Demented 1 0 M R 68 12 2 27 0,5 1457 0,806 1,205
Demented 1 0 F R 66 12 3 30 0,5 1447 0,769 1,213
Nondemented 1 0 F R 78 16 2 29 0 1333 0,748 1,316
Nondemented 1 0 F R 81 12 4 30 0 1230 0,715 1,427
Demented 1 0 M R 76 16 3 21 0,5 1602 0,697 1,096
Demented 1 0 M R 88 8 4 25 0,5 1651 0,66 1,063
Nondemented 1 0 M R 80 12 3 29 0 1783 0,752 0,985
Converted 1 0 F R 87 14 1 30 0 1406 0,715 1,248
Converted 1 0 M R 80 20 1 29 0 1587 0,693 1,106
Alzheimer’s disease is diagnosed based on the value of some key indicators.
Data analysis in table 3 illustrates some features with the average of each patients
(Demented and Non-Demented).
T a b l e 3 . Average of each features that could cause the Alzheimer’s disease
AVE Age EDUC SES MMSE CDR eTIV nWBV ASF
Female
Demented
76,10714 12,89286 3 25,46429 0,625 1373,929 0,728786 1,281036
Female
Non-Demented
75,68 15,22 2,38 29,34 0 1402,56 0,75072 1,26152
Male
Demented
74,33333 14,30556 2,5625 25,22222 0,583333 1555,278 0,720972 1,135361
Male
Non-Demented
74,86364 15,04545 2,5 28,86364 0 1656,364 0,735682 1,070909
Based on the table 3, it could be concluded that the average of education lev-
el for Demented people is less than Non-Demented. While average of the age for
Demented patients is higher than Non-Demented (for both female and male).
All features were compared separately during training Random Forest Algo-
rithm. After some experiments, Gender, Age and Education level were computed
that are the most valuable attributes for classification. The coefficient of impor-
tance each features for forest was multiplied by 100 which is illustrated in fig. 1.
There were used 70% of data for training and 30% of them for test. Accu-
racy of the classification using by Random Forest Algorithms was 89,5%.
Methods for improving accuracy of the dementia diagnosis using feature dimension reduction
Системні дослідження та інформаційні технології, 2019, № 2 29
Whereas, accuracy of the classification with Support Vector Machine was 88%.
The table 4 compares two algorithms result for classification.
T a b l e 4 . Average accuracy of each algorithms
Classifier Average Recall, % Average Precision, % Average F-Score, %
SVM 89 88 88
Random Forest 89 89 89
CONCLUSION
This article presents the results of known works analysis on diagnosis of Alz-
heimer’s disease. The following conclusions should be made after analysis of
considered works.
By using Random Forest Algorithms (RFA), we determine the most infor-
mative features like gender, age and education level which could mostly affect
Alzheimer’s disease.
In addition, Support vector machine and RF algorithms were compared with
accuracy 88% and 89% respectively.
In the future research, it’s planned to improve Convolutional Neural Net-
work to reach the higher accuracy. Based on previous works, CNN provides high
results in classification and has various benefits: like it’s very good feature extrac-
tors. In addition, the results of CNN classification will be compared with the algo-
rithms which were consider in the current paper.
REFERENCES
1. Convolutional Neural Network. 3 things you need to know. — P. 1–4. — Available
at: https://www.mathworks.com/solutions/deep-learning/convolutional-neural-
network.html
2. Альцгеймера болезнь. Диагностика. Анализы и инструментальные
исследования. — Режим доступа: http://demenciya.com
Feature importances of the forest
0
5
10
15
20
25
30
35
Age ASF CDR Educ eTIV M/F MMSE nWBV SES
Importance of each features
Maryam Naderan, Yuriy Zaychenko
ISSN 1681–6048 System Research & Information Technologies, 2019, № 2 30
3. Islam J. An Ensemble of Deep Convolutional Neural Networks for Alzheimer's Dis-
ease Detection and Classification, Computer Vision and Pattern Recognition /
J. Islam, Y. Zhang. — Available at: https://arxiv.org/pdf/1712.01675.pdf
4. Sarraf S. Deepad: Alzheimer’s disease classification via deep convolutional neural
networks using mri and fmri / S. Sarraf, J. Anderson, G. Tofighi. — bioRxiv,
p. 070441, 2016.
5. Naderan M. Diagnosing Lung Cancer Based on Deep Learning Algorithms: Review
/ M. Naderan, Y.P. Zaychenko // 20-th International conference on System
Analysis and Information Technology SAIT 2018, May 21–24, 2018. —
P. 111–112.
6. Satoki S. Application of Convolutional Neural Networks in the Diagnosis of Hel-
icobacter pylori Infection Based on Endoscopic Images / S. Satoki, N. Shuhei,
A. Kazuharu, N. Yoshitaka et al. // EBioMedicine. — Vol. 25, November 2017.
— P. 106–111.
7. Boysen Jacob. Magnetic Resonance Imaging Comparisons of Demented and Non-
demented Adults / Jacob Boysen. — Available at: https://www.kaggle.com/
jboysen/mri-and-alzheimers
Received 14.09.2018
From the Editorial Board: the article corresponds completely to submitted manuscript.
|
| id | journaliasakpiua-article-175552 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:26:04Z |
| publishDate | 2019 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/a1/2df96b89b846d8b7cabd3d01059e39a1.pdf |
| spelling | journaliasakpiua-article-1755522019-08-27T22:12:50Z Methods for improving accuracy of the dementia diagnosis using feature dimension reduction Методы улучшения точности диагностики деменции при помощи сокращения размерности признаков Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак Naderan, Maryam Zaychenko, Yuriy P. diagnosis Alzheimer’s disease ensemble learning methods classification convolutional neural network диагноз болезни Альцгеймера ансамблевые методы обучения классификация сверточная нейронная сеть діагноз хвороби Альцгеймера ансамблеві методи навчання класифікація згорткова нейронна мережа In this paper, the problem of choosing the right feature for diagnosing Dementia is discussed. Several features that could affect dementia were reviewed and their importance was evaluated. Random forest algorithm and SVM for the dementia diagnosis have been developed and investigated. Experiments were conducted on the open-source database and compared with the related works’ results. The purpose of the paper is to improve the accuracy of diagnosis of dementia using the reduction of features' dimension. This article is devoted to analysis of the main distinguishing features of Alzheimer`s dementia, applicable methods and treatment of Alzheimer's dementia on early stage that could help to avoid negative consequences connected with progress of the disease. The purpose of the paper is to improve the accuracy of diagnosis of dementia. Рассмотрена проблема выбора набора признаков для диагностики деменции. Предложен набор признаков для поставленной проблемы и оценена важность каждого из них. Разработан и исследован алгоритм на основе ансамбля деревьев решений и метода опорных векторов для диагностики деменции. Экспериментальные исследования проведены на основе базы данных, представленной в Kaggle. Выполнен сравнительный анализ полученных результатов с результатами существующих работ. Проанализированы основные отличительные признаки деменции Альцгеймера, применяемых методов лечения деменции Альцгеймера на ранней стадии, которые помогут избежать негативных последствий, связанных с прогрессированием заболевания. Целью работы является повышение точности диагностики деменции. Розглянуто проблему вибору набору ознак для діагностики деменції. Запропоновано набір ознак для поставленої проблеми й оцінено значущість кожної з них. Розроблено та досліджено алгоритм на основі ансамблю дерев рішень і методу опорних векторів для діагностики деменції. Експериментальні дослідження проведено на основі бази даних, поданої в Kaggle. Виконано порівняльний аналіз отриманих результатів з результатами наявних праць. Проаналізовано основні відмітні ознаки деменції Альцгеймера, застосовуваних методів лікування деменції Альцгеймера на ранній стадії, які можуть допомогти уникнути негативних наслідків, зумовлених прогресуванням захворювання. Метою роботи є підвищення точності діагностики деменції. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2019-06-25 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/175552 10.20535/SRIT.2308-8893.2019.2.03 System research and information technologies; No. 2 (2019); 25-30 Системные исследования и информационные технологии; № 2 (2019); 25-30 Системні дослідження та інформаційні технології; № 2 (2019); 25-30 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/175552/175461 Copyright (c) 2021 System research and information technologies |
| spellingShingle | діагноз хвороби Альцгеймера ансамблеві методи навчання класифікація згорткова нейронна мережа Naderan, Maryam Zaychenko, Yuriy P. Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title | Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title_alt | Methods for improving accuracy of the dementia diagnosis using feature dimension reduction Методы улучшения точности диагностики деменции при помощи сокращения размерности признаков |
| title_full | Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title_fullStr | Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title_full_unstemmed | Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title_short | Методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| title_sort | методи поліпшення точності діагностики деменції за допомогою скорочення розмірності ознак |
| topic | діагноз хвороби Альцгеймера ансамблеві методи навчання класифікація згорткова нейронна мережа |
| topic_facet | diagnosis Alzheimer’s disease ensemble learning methods classification convolutional neural network диагноз болезни Альцгеймера ансамблевые методы обучения классификация сверточная нейронная сеть діагноз хвороби Альцгеймера ансамблеві методи навчання класифікація згорткова нейронна мережа |
| url | https://journal.iasa.kpi.ua/article/view/175552 |
| work_keys_str_mv | AT naderanmaryam methodsforimprovingaccuracyofthedementiadiagnosisusingfeaturedimensionreduction AT zaychenkoyuriyp methodsforimprovingaccuracyofthedementiadiagnosisusingfeaturedimensionreduction AT naderanmaryam metodyulučšeniâtočnostidiagnostikidemenciipripomoŝisokraŝeniârazmernostipriznakov AT zaychenkoyuriyp metodyulučšeniâtočnostidiagnostikidemenciipripomoŝisokraŝeniârazmernostipriznakov AT naderanmaryam metodipolípšennâtočnostídíagnostikidemencíízadopomogoûskoročennârozmírnostíoznak AT zaychenkoyuriyp metodipolípšennâtočnostídíagnostikidemencíízadopomogoûskoročennârozmírnostíoznak |