Yazar "Taghiyev, Anar" seçeneğine göre listele

Listeleniyor 1 - 3 / 3
  • Küçük Resim Yok
    Öğe
    Advanced Image Processing Techniques and Applications for Biological Objects
    (IEEE, 2017) Altun, Adem Alpaslan; Taghiyev, Anar
    In this article, modern technical means of obtaining, processing and analysis of images for biological objects are studied. Whole process from recording the image, to processing it, is dissected both in hardware and technical details. Information about automated systems of processing for general purpose researched, and also confocal laser scanning microscope specialized system of image processing and techniques behind the analysis for micro objects are considered. The confocal approach made it much easier to obtain images of living microobjects, made it possible to automate the accumulation of three-dimensional data, and improved images of drugs using simultaneously several fluorescent markers.
  • Küçük Resim Yok
    Öğe
    A Machine Learning Framework to Identify the Causes of HbA1c in Patients With Type 2 Diabetes Mellitus
    (ROMANIAN SOC CONTROL TECH INFORMATICS, 2019) Taghiyev, Anar; Altun, Adem Alpaslan; Allahverdi, Novruz; Çağlar, Sona
    In this study, the effects of blood glucose levels on hemoglobin A1c (HbA1c) were investigated. For this reason, a classification model was developed by carrying out a logistic regression analysis based on machine learning and data mining methods. The purpose of using logistic regression analysis in this study was to establish a method of creating a statistical model that is most suitable and reasonable for determining the relationship between dependent and independent variables. This model shows how effective the factors that cause an increase in the HbA1c level. It can be planned to verify this method on more Electronic Heath Records databases to address the learning method of information in the local health sector with the help of data mining and machine learning methods and different clinical problems for future work.
  • Küçük Resim
    Öğe
    Medical data analysis and model development based on machine learning using apache spark technology
    (Selçuk Üniversitesi Fen Bilimleri Enstitüsü, 2021) Taghiyev, Anar; Altun, Adem Alpaslan
    This study focuses on how to use machine-learning methods (based on Apache Spark technology) to create value for healthcare institutions and define strategies for applying big data technology. In the thesis, an attempt was made to describe the current state of big data technology, to give a general perspective, and to emphasize the main goals and objectives of the application of Apache Spark technology in healthcare. In other words, this study explores the trajectory of the scientific and technological development of the techniques used to analyze data in healthcare. Inspired by this idea, an attempt was made to use the logistic regression method to identify the causes of diseases using the Apache Spark technology, which is the main purpose of the study. In the first phase of the study, a logistic regression analysis was performed using a dataset available online in the UCI Machine Learning Repository, and a new single-stage model was developed to identify the causes of HbA1c in patients with type 2 diabetes. The purpose of using logistic regression analysis in this study was to establish a method for creating a statistical model that is most appropriate and reasonable for determining the relationship between dependent and independent variables. The independent variables had continuous and categorical values in the classification model, and binary interactions of independent variables were included as a common variable. The interpretations of the results obtained by practical examples are also described in detail. The single-stage classification model developed based on logistic regression analysis, showed a more accurate and effective approach for patients with poorly controlled diabetes and well-controlled diabetes. In the second phase of the study, an attempt was made to conduct a questionnaire survey in the region of Turkey in females aged 18 years and above, as well as to extract medical data of our study participants from the database of Electronic Health Records of Aksaray Sultanhani Family Health Center, and then to conduct a logistic regression analysis to identify the causes of obesity in patients. The motivation for studying obesity in the region of Turkey was that the obesity issue has international relevance in recent years. Because referring to the data of the World Health Organization, approximately 600 million (13%) adults aged 18 years and above are obese; at least 2.8 million people die annually from obesity and overweight. In Turkey, the prevalence of obesity is 20.5 percent among adult men and 41 percent among women. The aim of this study is to develop a hybrid model to identify the causes of obesity in the region of Turkey in females aged 18 years and above. The data used in this study was collected from Aksaray Sultanhani Family Health Center, in the period from March to November 2019, followed by a medical data analysis. Given the structure of medical data and questionnaires, an attempt was made to develop a two-stage model to achieve better results in identifying the causes of obesity in females aged 18 years and above. The first stage is the selection of features (i.e., the best variables) through the decision trees method, while the second stage is classification performed by the logistic regression method. The effectiveness of the proposed two-stage hybrid approach has been compared with traditional single-stage classifiers such as decision trees and logistic regression. In the thesis, the performance of the proposed single-stage and two-stage models have been was evaluated by two validation methods: namely holdout validation and five-fold cross-validation using the measurements, such as accuracy, specificity and sensitivity, precision, and Fmeasure. In conclusion, it should be emphasized that the proposed hybrid system gives %91 accuracy (mean), which is better than other single-stage classifiers. Thus, the proposed hybrid system provides a more accurate classification of obese patients and a practical approach to evaluating the factors affecting obesity. In this thesis, the final outputs/findings of the proposed method are to get the odds ratio of the more effective factors that cause of obesity that is discussed in detail. According to the findings, women should be provided with training and counseling on the risks of obesity and possible health problems. Women should be informed about the weight they should gain during pregnancy, how to lose weight after pregnancy, and recommendations should be related to nutrition and physical activity. In order to protect, maintain, and improve health, it is important to control the factors that affecting obesity and to prepare health education programs for primary healthcare professionals (as well as experts/specialists). Primary healthcare professionals should be better informed about these issues.