Application of deep learning methods for automated analysis of retinal structures in ophthalmology

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 14, No. 2, April 2024, pp. 1987~1995
ISSN: 2088-8708, DOI: 10.11591/ijece.v14i2.pp1987-1995  1987
Journal homepage: http://ijece.iaescore.com
Application of deep learning methods for automated analysis of
retinal structures in ophthalmology
Akmaral Kassymova1
, Assem Konyrkhanova2
, Aida Issembayeva1
, Zagira Saimanova3
,
Alisher Saltayev4
, Maral Ongarbayeva5
, Gulnur Issakova6
1
Department of Information Technology, Zhangir Khan University, Uralsk, Republic of Kazakhstan
2
Department of Information Security, Faculty of Information Technologies, L.N. Gumilyov Eurasian National University,
Astana, Republic of Kazakhstan
3
Department of Information and Computing System, Abylkas Saginov Karaganda Technical University, Karaganda,
Republic of Kazakhstan
4
Faculty of Natural Sciences, L.N. Gumilyov Eurasian National University, Astana, Republic of Kazakhstan
5
Department of Information and Communication Technologies, International Taraz Institute named after Sherkhan Murtaza,
Taraz, Republic of Kazakhstan
6
Department of Information Systems, Kazakh Agrotechnical Research University named after S. Seifullin, Astana,
Republic of Kazakhstan
Article Info ABSTRACT
Article history:
Received Oct 8, 2023
Revised Dec 1, 2023
Accepted Dec 13, 2023
This article examines a current area of research in the field of ophthalmology
the use of deep learning methods for automated analysis of retinal structures.
This work explores the use of deep learning methods such as EfficientNet
and DenseNet for the automated analysis of retinal structures in
ophthalmology. EfficientNet, originally proposed to balance between
accuracy and computational efficiency, and DenseNet, based on dense
connections between layers, are considered as tools for identifying and
classifying retina features. Automated analysis includes identifying
pathologies, assessing the degree of their development and, possibly,
diagnosing various eye diseases. Experiments are performed on a dataset
containing a variety of images of retinal structures. Results are evaluated
using metrics of accuracy, sensitivity, and specificity. It is expected that the
proposed deep learning methods can significantly improve the automated
analysis of retinal images, which is important for the diagnosis and
monitoring of eye diseases. As a result, the article highlights the significance
and promise of using deep learning methods in ophthalmology for
automated analysis of retinal structures. These methods help improve the
early diagnosis, treatment and monitoring of eye diseases, which can
ultimately lead to improved healthcare quality and improved patient lives.
Keywords:
Deep learning
DenseNet
EfficientNet
Eye diseases
Ophthalmology
Pathology
This is an open access article under the CC BY-SA license.
Corresponding Author:
Assem Konyrkhanova
Department of Information Security, Faculty of Information Technologies, L.N. Gumilyov Eurasian National
University
010000 Astana, Republic of Kazakhstan
Email: konyrkhanova@internet.ru
1. INTRODUCTION
In modern ophthalmology, advanced deep learning technologies are being actively implemented [1],
[2] for automated analysis of retinal structures. Some of the key methods that have attracted the attention of
researchers and clinicians are EfficientNet [3], [4] and DenseNet [5], [6]. These methods provide powerful
tools for image processing [7], [8] with a high degree of efficiency. The retina, as a key element of the visual

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 14, No. 2, April 2024: 1987-1995
1988
system, is an object of attention in medical diagnostics. Assessing the condition of the retina and identifying
pathologies requires accuracy and high sensitivity. In this context, deep learning methods [9], [10] provide a
unique opportunity for automated analysis of retinal structures [11] with a high degree of accuracy.
EfficientNet, originally designed to achieve an optimal balance between accuracy and computational
efficiency, provides efficient neural network architectures [12]–[14]. On the other hand, DenseNet [15],
based on dense connections, can efficiently extract features from images while preserving information at all
levels. In this study, we focus on the application of EfficientNet and DenseNet for automated analysis of
retinal structures [16]. These methods can significantly improve the processes of identifying and classifying
pathologies [17], which in turn will facilitate the diagnosis and monitoring of eye diseases. Our work aims to
investigate the effectiveness of these methods on a variety of data sets, as well as to evaluate their
applicability in the clinical practice of ophthalmology The results of this study have the potential to
significantly improve the accuracy and speed of diagnosis of ophthalmic conditions, opening new prospects
for improving patient vision care. Viedma et al. [18] discusses methods for automatic analysis of retinal
optical coherence tomography (OCT) images using deep learning (DL). The authors highlight the
fundamental importance of such analysis in providing quantitative data on posterior ocular health and discuss
the advantages of DL over traditional methods. The paper presents a detailed literature review, covering
state-of-the-art DL-based methods for segmenting retinal layers in OCT images. The publication results
highlight the success of DL in ocular image analysis, providing high performance and improved accuracy
compared to previous machine learning methods and traditional image analysis. The authors also highlight
promising directions for future research in this area.
Zekavat et al. [19] consider the importance of the microvascular network, especially in the fundus of
the retina, as a key element in maintaining organ health and tumor development. They highlight that assessing
this microcirculatory system using machine learning can provide new research opportunities for understanding
human health and disease. The authors focus on the importance of whole-phenomenal and genome-wide
analysis of the fundus vasculature and propose a large-scale assessment that can bring new insights into medical
diagnosis and therapy. Jeong et al. [20] discuss the importance of automation in the field of ophthalmology,
especially in the analysis of fundus images for the detection and diagnosis of various conditions such as diabetic
retinopathy, age-related macular degeneration and glaucoma. They emphasize that the use of deep learning and
machine learning techniques allows machines to efficiently interpret and analyze complex medical data, which
can save doctors time and reduce the likelihood of diagnostic errors. The article provides an overview of recent
research in this area, including screening and diagnostic methods, and retinal vasculature extraction from fundus
images, with an emphasis on the challenges these systems face.
Kumar and Gupta [21] proposed an efficient hybrid deep learning method for segmentation and
classification of retinal blood vessels. The images are pre-processed to improve quality, then a new enhanced
fuzzy c-means (EFCM) clustering scheme is used to highlight important segments. Grouping images by vessel
thickness helps reduce computational complexity. Experiments on digital retinal images for vessel extraction
(DRIVE), structured analysis of the retina (STARE) and high-resolution fundus (HRF) databases show that the
proposed model outperforms state-of-the-art methods, achieving high accuracy rates (99% for DRIVE, 98% for
STARE and HRF). Abbood et al. [22] focus on the importance of early diagnosis of diabetic retinopathy (DR)
to prevent vision loss in patients with diabetes. They present a fundus image enhancement algorithm used to
improve the clarity and contrast of retinal photographs. The algorithm includes steps of cropping images and
applying Gaussian blur to reduce noise. Experimental results conducted on EyePACS and MESSIDOR datasets
show significant improvement in feature extraction and classification on enhanced images. The improved
algorithm has also been successfully tested in smart hospitals as an internet of medical things (IoMT)
application, demonstrating its potential for practical application in medical practice. Zhdanov et al. [23] present
a method for diagnosing retinal dystrophy using electroretinography and develop an algorithm to support
medical decision-making for an accurate diagnosis. Using machine learning methods, the algorithm is based on
parameters extracted from wavelet scalograms of electroretinogram signals in patients of different ages. The
study involves the use of a labeled database of electroretinogram signals from children and adults. The scientific
novelty of the work lies in the development of special mathematical and algorithmic software for analyzing the
parameters of the wavelet scalogram of the electroretinogram signal. The proposed algorithm also provides a
more accurate classification of electroretinogram signals in adults and children compared to the classical
analysis method.
2. METHOD
Research and analysis of the retina using deep learning methods, in this case, focuses on multi-class
classification, also known as “muti label classification” [24]. This approach [25] assumes that each retinal
image can be classified into several classes corresponding to different structures or states of the eye. In the

Int J Elec & Comp Eng ISSN: 2088-8708 
Application of deep learning methods for automated analysis of retinal … (Akmaral Kassymova)
1989
context of ophthalmology, this may involve the classification of various diseases, changes or features of the
retinal structures. Examples of classes include conditions such as macular degeneration, glaucoma, diabetic
retinopathy, and normal retina. Multi-class classification using deep learning allows for more accurate and
automated detection of a variety of retinal-based eye conditions. This is of great importance for early
diagnosis and monitoring of eye diseases, which, in turn, helps to improve the quality of ophthalmic care.
The EfficientNetB0 architecture is part of an efficient family of models designed to optimize model
accuracy and size. The strategy relies on complex scaling of three dimensions of the network, including
width, depth and resolution, using a complex scaling factor. EfficientNetB0 is the base model in this family,
serving as a starting point for the development of larger and more powerful variations. As shown in Figure 1,
unlike traditional approaches that are limited to scaling only one dimension, EfficientNet simultaneously
scales all three dimensions, which helps achieve higher accuracy with comparable or even smaller model
size. DenseNet, in turn, stands out for its feature of “dense” connections between layers, where each layer
receives input from all previous layers. Unlike DenseNet, EfficientNet does not use such a dense connection
structure, which leads to differences in the number of parameters. DenseNet169 typically has more
parameters due to its dense connectivity, which can increase model size and inference time. In terms of
computational efficiency, EfficientNetB0 generally requires less computational resources to achieve
comparable or even better performance compared to DenseNet169. On a number of tasks and datasets,
EfficientNet can outperform DenseNet in accuracy with a lower number of parameters. The EfficientNetB0
model offers a modern approach to scaling neural networks, providing high performance in a comparatively
smaller size compared to other architectures such as DenseNet169. The choice between these models should
be based on the specific requirements of the task, available computing resources and data characteristics.
The application of deep learning methods such as EfficientNet and DenseNet in the field of
ophthalmology is becoming a key step in automating the analysis of retinal structures. These innovative
methods reduce the dependence on human error, providing faster and more accurate diagnosis of various eye
conditions. Ultimately, such technologies offer promise for improving the quality of medical care and
optimizing treatment strategies in ophthalmology.
Figure 1. Architecture of deep learning models

 ISSN: 2088-8708
1990
3. RESULTS AND DISCUSSION
In our research, we successfully implemented deep learning methods for automated analysis of
retinal structures in ophthalmology. Our results using the EfficientNet and DenseNet models confirmed
significant advances in the field of automated analysis of ocular structures. In the context of our results, it is
also important to discuss potential limitations and possible directions for future research. We see prospects
for optimizing algorithms, expanding datasets to take into account the diversity of pathologies, and carefully
adapting models to different patient groups. Thus, the use of deep learning, in particular EfficientNet and
DenseNet, in the field of ophthalmology provides significant opportunities for automating and improving the
processes of diagnosing eye diseases. As a result of the study, the performance of machine learning models,
namely DenseNet169 and EfficientNetB0, was compared at the training and validation stages. The training
and validation losses for DenseNet169 and EfficientNetB0 show high initial values, followed by a stable
decrease and different final values. DenseNet169 shows a decreasing trend in loss over time, reaching a
plateau, while EfficientNetB0 also shows a stable decrease in loss throughout training. Regarding accuracy,
both training methods, DenseNet169 and EfficientNetB0, show increasing values over time as shown in
Figure 2. However, EfficientNetB0 stands out for its higher accuracy value in the validation phase compared
to DenseNet169. The final results indicate that EfficientNetB0 outperforms DenseNet169 in the considered
metrics, exhibiting lower final loss and higher accuracy in the validation phase. Specifically, EfficientNetB0
exhibited a final training loss of 0.1152 and a validation loss of 0.2025, with a corresponding final training
accuracy of 95.66% and validation of 92.77%. While DenseNet169 has final training and validation losses of
0.1811 and 0.2582, respectively, with a final training accuracy of 93.32% and validation accuracy of 91.12%
as shown in Figure 2(a). Thus, based on the presented data, in Figure 2(b) the EfficientNetB0 model shows
more effective results in comparison with DenseNet169, which confirms its advantages in the context of the
considered performance metrics.
In this study, we conducted an in-depth analysis of retinal images to identify and classify various
diseases. The retina is a unique structure of the eye that plays a key role in vision, and diseases of the retina
can lead to decreased or loss of vision. Thus, rapid and accurate diagnosis of retinal diseases is a key aspect
in ophthalmology. The total training sample consisted of 1920 images, which represented various retinal
diseases. From the entire variety of possible classes of diseases, those that were represented in the largest
number (more than 100 images for each class) were selected to provide a sufficient amount of data for
training and validating the models. As a result, the following five classes were selected: i) diabetic retinopathy
(DR) with 376 images, ii) macular hole (MH) with 317 images, iii) diabetic neuropathy (DN) with 138 images,
iv) type-specific lymph node (TSLN) with 186 images, and v) optic disc cupping (ODC) with 282 images.
Two advanced deep learning architectures were used for image classification: DenseNet169 and
EfficientNetB0. These architectures were selected based on their superior performance in a number of computer
vision competitions and studies. After training and validation on retinal data, EfficientNetB0 performed better
than DenseNet169 in terms of loss and accuracy during the validation phase as shown in Figure 3.
On the other hand, DenseNet, although it performed well, was slightly inferior to EfficientNet in
some metrics. However, this method also provided high performance, making it a significant tool for
automated analysis of retinal structures. It is important to note that the performance of each model may
depend on the specific study conditions, characteristics of the data used, and diagnostic purposes. Thus, as
shown in Figure 4, the choice between EfficientNet and DenseNet can be justified by the context of a
particular task and the requirements of a particular clinical scenario.
By analyzing the loss and accuracy plots for both models, the loss plot for training and validation
shows a steady decrease over time, indicating successful training. The initial loss is much higher, and
although it also decreases over time, it remains higher compared to binary classification. The accuracy of
training for binary classification and validation increases steadily, reaching high values. And the training
accuracy of multi-class classification and validation also increases but remains lower compared to binary
classification. Binary classification basically distinguishes between two classes or states, making the task
relatively simple compared to multi-class classification, where the model must distinguish between many
classes. In multi-class classification, the feature space becomes much more complex, and the model requires
more training time and data to achieve similar performance to binary classifiers. In addition, with multiclass
classification, the probability of random guessing is much lower than with binary classification, which can
also affect the initial metrics of the model. In the case of retinal images, different diseases may have similar
visual features, making the task of multi-class classification even more challenging. Effective learning in
such environments requires more complex models, additional data, and possibly the use of data augmentation
or transfer learning techniques. In conclusion, differences in performance between binary and multiclass
classification can be explained by task complexity, data volume and quality, and model architecture features.
This highlights the importance of choosing an appropriate model and learning strategy depending on the
specific task and available data.

1991
(a)
(b)
Figure 2. Accuracy metrics of deep learning models (a) according to the DenseNet169 model and
(b) according to the EfficientNet model

 ISSN: 2088-8708
1992
Figure 3. Comparative analysis of the performance of deep learning methods
Figure 4. Comparative analysis of binary and multiclass classification indicators
4. CONCLUSION
In conclusion, a study on the application of deep learning methods EfficientNet and DenseNet for
automated analysis of retinal structures in ophthalmology revealed important conclusions. The EfficientNet

1993
method demonstrated impressive high accuracy in the task of analyzing retinal structures. The results
obtained indicate its effectiveness and potential applicability in clinical practice. The effectiveness of
EfficientNet, expressed in high classification accuracy values, highlights its ability to reliably and accurately
recognize retinal pathologies. These findings may be important for improving diagnostic methods in
ophthalmology, as well as for increasing the efficiency and efficiency of medical decisions. Our research
demonstrates the potential of deep learning to improve diagnostic processes for ocular diseases, reducing
reliance on human error and increasing the overall accuracy of retinal structure analysis. These methods not
only speed up the diagnostic process, but also promise more accurate and reliable results. An important
aspect of our work is the recognition of the need for further research in this area, including optimization of
algorithms, expansion of datasets and adaptation of models to different patient groups. This opens up the
prospect of creating more comprehensive decision support systems in ophthalmology, which could
revolutionize the diagnosis and treatment of eye diseases.
REFERENCES
[1] S. M. Panchal and Shivaputra, “Precise identification of objects in a hyperspectral image by characterizing the distribution of pure
signatures,” International Journal of Electrical and Computer Engineering (IJECE), vol. 12, no. 6, pp. 6068–6078, 2022, doi:
10.11591/ijece.v12i6.pp6068-6078.
[2] M. C. G. Baabu and P. M. C., “Semantic feature extraction method for hyperspectral crop classification,” Indonesian Journal of
Electrical Engineering and Computer Science (IJEECS), vol. 23, no. 1, pp. 387–395, 2021, doi: 10.11591/ijeecs.v23.i1.pp387-395.
[3] V. V. G. and T. K, “Spatial-spectral feature for extraction technique for hyperspectral crop classification,” Indian Journal of
Science and Technology, vol. 15, no. 2, pp. 81–90, 2022, doi: 10.17485/ijst/v15i2.1810.
[4] K. Zhang et al., “Panchromatic and multispectral image fusion for remote sensing and earth observation: concepts, taxonomy,
literature review, evaluation methodologies and challenges ahead,” Information Fusion, vol. 93, pp. 227–242, May 2023, doi:
10.1016/j.inffus.2022.12.026.
[5] M. C. G. Babu and P. M. C., “Inherent feature extraction and soft margin decision boundary optimization technique for
hyperspectral crop classification,” International Journal of Advanced Computer Science and Applications, vol. 12, no. 12, 2021,
doi: 10.14569/IJACSA.2021.0121285.
[6] A. Asokan, J. Anitha, B. Patrut, D. Danciulescu, and D. J. Hemanth, “Deep feature extraction and feature fusion for bi-temporal
satellite image classification,” Computers, Materials and Continua, vol. 66, no. 1, pp. 373–388, 2021, doi:
10.32604/cmc.2020.012364.
[7] O. Tamin, E. G. Moung, J. A. Dargham, F. Yahya, and S. Omatu, “A review of hyperspectral imaging-based plastic waste
detection state-of-the-arts,” International Journal of Electrical and Computer Engineering (IJECE), vol. 13, no. 3,
pp. 3407–3419, Jun. 2023, doi: 10.11591/ijece.v13i3.pp3407-3419.
[8] C. H. Feng, Y. Makino, S. Oshita, and J. F. García Martín, “Hyperspectral imaging and multispectral imaging as the novel
techniques for detecting defects in raw and processed meat products: Current state-of-the-art research advances,” Food Control,
vol. 84, pp. 165–176, 2018, doi: 10.1016/j.foodcont.2017.07.013.
[9] F. Behrad and M. Saniee Abadeh, “An overview of deep learning methods for multimodal medical data mining,” Expert Systems
with Applications, vol. 200, p. 117006, Aug. 2022, doi: 10.1016/j.eswa.2022.117006.
[10] M. Tsuneki, “Deep learning models in medical image analysis,” Journal of Oral Biosciences, vol. 64, no. 3, pp. 312–320, Sep.
2022, doi: 10.1016/j.job.2022.03.003.
[11] G. B. Abdikerimova et al., “Analysis of images obtained by the transmission electronic microscopy,” Materials Today:
Proceedings, vol. 12, pp. 90–92, 2019, doi: 10.1016/j.matpr.2019.03.071.
[12] N. A. Andriyanov, V. E. Dementiev, and A. G. Tashlinskiy, “Detection of objects in the images: from likelihood relationships
towards scalable and efficient neural networks,” Computer Optics, vol. 46, no. 1, Feb. 2022, doi: 10.18287/2412-6179-co-922.
[13] Z. Han, M. Jian, and G.-G. Wang, “ConvUNeXt: an efficient convolution neural network for medical image segmentation,”
Knowledge-Based Systems, vol. 253, Oct. 2022, doi: 10.1016/j.knosys.2022.109512.
[14] M. A. Abdou, “Literature review: efficient deep neural networks techniques for medical image analysis,” Neural Computing and
Applications, vol. 34, no. 8, pp. 5791–5812, Feb. 2022, doi: 10.1007/s00521-022-06960-9.
[15] C. Gong, J. Liu, M. Gong, J. Li, U. A. Bhatti, and J. Ma, “Robust medical zero‐watermarking algorithm based on Residual‐
DenseNet,” IET Biometrics, vol. 11, no. 6, pp. 547–556, Sep. 2022, doi: 10.1049/bme2.12100.
[16] J. Freiberg et al., “Automated analysis of vessel morphometry in retinal images from a Danish high street optician setting,” PLOS
ONE, vol. 18, no. 8, Aug. 2023, doi: 10.1371/journal.pone.0290278.
[17] G. Abdikerimova et al., “Detection of chest pathologies using autocorrelation functions,” International Journal of Electrical and
Computer Engineering (IJECE), vol. 13, no. 4, pp. 4526–4534, Aug. 2023, doi: 10.11591/ijece.v13i4.pp4526-4534.
[18] I. A. Viedma, D. Alonso-Caneiro, S. A. Read, and M. J. Collins, “Deep learning in retinal optical coherence tomography (OCT): a
comprehensive survey,” Neurocomputing, vol. 507, pp. 247–264, Oct. 2022, doi: 10.1016/j.neucom.2022.08.021.
[19] S. M. Zekavat et al., “Deep learning of the retina enables phenome- and genome-wide analyses of the microvasculature,”
Circulation, vol. 145, no. 2, pp. 134–150, Jan. 2022, doi: 10.1161/circulationaha.121.057709.
[20] Y. Jeong, Y.-J. Hong, and J.-H. Han, “Review of machine learning applications using retinal fundus images,” Diagnostics, vol.
12, no. 1, Jan. 2022, doi: 10.3390/diagnostics12010134.
[21] Y. Kumar and B. Gupta, “Retinal image blood vessel classification using hybrid deep learning in cataract diseased fundus
images,” Biomedical Signal Processing and Control, vol. 84, Jul. 2023, doi: 10.1016/j.bspc.2023.104776.
[22] S. H. Abbood, H. N. A. Hamed, M. S. M. Rahim, A. Rehman, T. Saba, and S. A. Bahaj, “Hybrid retinal image enhancement
algorithm for diabetic retinopathy diagnostic using deep learning model,” IEEE Access, vol. 10, pp. 73079–73086, 2022, doi:
10.1109/access.2022.3189374.
[23] A. E. Zhdanov, A. Y. Dolganov, D. Zanca, V. I. Borisov, E. Luchian, and L. G. Dorosinsky, “Evaluation of the effectiveness of
the decision support algorithm for physicians in retinal dystrophy using machine learning methods,” Computer Optics, vol. 47,
no. 2, Apr. 2023, doi: 10.18287/2412-6179-co-1124.
[24] T. Pham, X. Tao, J. Zhang, J. Yong, Y. Li, and H. Xie, “Graph-based multi-label disease prediction model learning from medical
data and domain knowledge,” Knowledge-Based Systems, vol. 235, Jan. 2022, doi: 10.1016/j.knosys.2021.107662.

 ISSN: 2088-8708
1994
[25] Q. Chen et al., “Multi-label classification for biomedical literature: an overview of the BioCreative VII LitCovid Track for
COVID-19 literature topic annotations,” Database, vol. 2022, Jan. 2022, doi: 10.1093/database/baac069.
BIOGRAPHIES OF AUTHORS
Akmaral Kassymova candidate of Pedagogical Sciences, currently works at
Zhangir Khan University at the Department of Information Technology, Professor. He has
more than 39 years of scientific and pedagogical experience and more than 100 scientific
papers, including 9 articles in the Scopus database, 20 teaching aids. She can be contacted at
email: kasimova_ah@mail.ru.
Assem Konyrkhanova has a scientific and pedagogical experience of more than
25 years, acting associate professor of the Department of Information Security of the Faculty
of Information Technologies of the L.N. Gumilyov ENU. Graduated from EKSU, Faculty of
Theoretical and Applied Mathematics, postgraduate studies in the specialization “Algebra and
mathematical Logic”, master's and doctoral studies in the specialty “Mathematics” at
D.Serikbayev EKSTU, Ust-Kamenogorsk. She defended her PhD thesis in the field of group
theory and computability. She has more than 40 scientific papers, including 2 articles based on
Scopus, 6 textbooks with ISBN, 4 electronic textbooks, 10 methodological manuals and
developments, and is the author of a certificate on entering information into the state register
of rights to objects protected by copyright: “Software for analyzing the mathematical model of
the development of ethnic groups of Kazakhstan”, type of copyright object: computer
program. She can be contacted at email: konyrkhanova@internet.ru.
Aida Issembayeva graduated from the Kazakh National Technical University
named after Satpayev in 2008 with a degree in automation and informatization in control
systems. In 2013, she graduated from the master’s degree program of Turan University with a
degree in computer engineering and software and received a master’s degree in technical
sciences. Currently, he is a senior lecturer at Zhangirkhan Zkatu. She is the author of more
than 10 articles on scientific and technical areas. She can be contacted at email:
aida_bz@mail.ru.
Zagira Saimanova received the academic degree of Master of Applied
Mathematics in the specialty 510251 “Applied Mathematics”, and the degree of Doctor of
Philosophy in the specialty 6D070400 – “Computer Engineering and Software”, Currently
works Abylkas Saginov Karaganda Technical University. She is the author and co-author of
more than 20 publications. The h-index is 1. Her research interests are in the field of
information technology and software development of computer technology and information
systems. She can be contacted at email: zagira_sb@mail.ru.
Alisher Saltayev graduated from the S.Seifullin Kazakh Research Agro
Technical University in 2023 with a degree in information systems and received a master's
degree in technical sciences. Currently, he is the head of the educational laboratory center of
the Faculty of Natural Sciences of the L.N. Gumilyov Eurasian National University. He is the
author of 2 scientific articles. Research interests in big data, computer analysis of DNA, RNA
and protein sequences, theoretical and computer analysis of structures, data mining, functional
analysis of genomes and proteins. He can be contacted at email: alisherskg@gmail.com.

1995
Maral Ongarbayeva associate professor of the Department of Information and
Communication Technologies of the International Taraz Institute named after Sherkhan
Murtaza, Republic of Kazakhstan. In 2002, she received the degree of Candidate of
Pedagogical Sciences in the specialty 13.00.02 - Theory and methodology of teaching and
upbringing (mathematics). She is engaged in research on the issues of informatization of
education, methods of teaching mathematics and computer science. She is a participant in the
project of grant financing of the Republic of Kazakhstan “Research and development of an
automated proctoring system for monitoring students' knowledge in distance learning” and the
project “computational life sciences” within the framework of the Erasmus program. She can
be contacted at email: ongarbayevamaral0@gmail.com.
Gulnur Issakova in 2003 she graduated from the South Kazakhstan State
University named after M. Auezov with a degree in mathematics and computer science. In
2022, she defended her Ph.D. dissertation in the specialty “6D012000 – Professional training”
and received a Ph.D. degree. Currently, she is a senior lecturer, Ph.D., Department of
Information Systems, Kazakh Agrotechnical Research University named after S. Seifullin. She
is the author of more than 50 scientific works, including 1 monograph and 3 articles in the
Scopus database. Scientific interests: image processing, pattern recognition theory, data
mining, geographic information systems. She can be contacted at email: is_gul_oral@mail.ru.

Application of deep learning methods for automated analysis of retinal structures in ophthalmology

More Related Content

Similar to Application of deep learning methods for automated analysis of retinal structures in ophthalmology (20)

More from IJECEIAES (20)

Recently uploaded (20)

Application of deep learning methods for automated analysis of retinal structures in ophthalmology