A Study of Designing a Deep Learning Model for Diabetic Retinopathy Diagnosis
 
Anuja Singh1*, Dr. Keshav Samrat Modi2
1 Research Scholar, Shri Krishna University, Chhatarpur M.P.
2 Assistant Professor, Shri Krishna University, Chhatarpur M.P.
Abstract - The healthcare industry is not immune to the widespread changes brought about by artificial intelligence and machine learning (AI/ML). Artificial intelligence and machine learning have the ability to greatly expand access to diabetes treatment, which would improve efficiency. With such a high prevalence of diabetes, India poses a unique combination of challenges, but also presents an interesting opportunity in terms of the data that may be made available to study the disease. The data for our study was gathered from a variety of hospitals. Physicians who have been in practice for more than 5 years are chosen as responders. For these two class problem, researcher define specificity as the number of images correctly identified as diabetic retina (diabretina) and normal retina (normalretina). Researcher defines accuracy in the form of the images of retina with a accurate classification. This final designed network succeeded with 80% of accuracy.
Keyword - machine learning, Artificial intelligence
INTRODUCTION
In recent years, the future increases in population have been gradually driven by driven healthcare innovation and improved living circumstances and greater overall welfare. We are thus now faced with a significant proportion of people with chronic infections. AI may be described as ways for PCs to make orders requiring regular understanding of people. Diabetes mellitus is an endless, unavoidable disease, rich in information and with a variety of potential outcomes. This is a productive approach to combine artificial intelligence with diabetes. Some of the polls also revealed that the number of patients globally at approximately 600 million would be noted by the end of 2040. The objective retention of blood glucose (70–180 mg/dL) may help prevent confusions with hyperglycemia (>180 mg/dL) and diabetic (hypoglycemia). The diabetes board tries to detect one out of two diabetic adults but spends 10 per cent (USD 760 billion) on diabetes in the globe. Computerized argument finds general application in four important areas of diabetic care: clinical choice, precious definition of hazards in the population, robotic retinal screening and patient self-administration. A clinical expert or a programmed device can either identify the DM physically. Any of these kinds of DM estimates have benefits and disadvantages. The main desired position of manual decision is that no machine support for the DM placement strategy has to be taken into consideration, thus allowing a clinical expert to be an area authority. The secondary symptoms of DM are sometimes so small in its underlying phase that they cannot be fully distinguished even by an experienced expert. Due to developments in the ML and AI, a computerized program is more probable and more competent than the manual DM recognition method in the identification and diagnosis of illness at a basic level. Advantages include a reduced residual job for healthcare experts and a little risk of human deficiency. Choice of emotionally helpful networks based on the PC may do an energetic job for productive discoveries and benefits administration. The present essay looks at how AI is expected to alter diabetes treatment as a visible component of artificial intelligence.
DIABETES MELLITUS
The word mellitus describes a metabolism of a variety of etiologies represented by continuous hyperglycemia with agitated effects of sugar, fat and protein assimilation owing to insulin discharge defects, movement of insulin or two. Diabetes mellitus (DM) is a problem involving how the body transforms into vitality over meals. The consequences of mellitus diabetes reinforce long-term harm to different bodies. The manifestations of diabetes mellitus may vary, such as, thirst, polyuria, eyesight clouds and weight loss. DM consumes meals once, translates the body into glucose-called sugar and transfers it to the circulatory tract. The pancreas generates insulin, a hormone which uses blood glucose to the cells for its necessary use in big measure. Normally, there are indications that are or cannot be true and hyperglycemia in such a way may be accessible long before the examination is carried out, to make fanatical and utilitarian adjustments. The boy does not produce insulin as it does if you have DM and are not looking for medications. Glucose in particular suffers in the body, which is mostly termed high glucose. The variants of the starch, fat and protein assimilation standard are due to the lack of insulin movements on target tissues resulting from brutality or insulin disregard. DM generates according to the source in various ways. The consequences from diabetes mellitus solidify dynamic growth in particular complexity, with predicted visual weakening, nephropathy that may cause renal deception, or maybe neuropathy, with danger of foot ulcers, evacuation, Charcot joints and autonomous characteristics, including sexual intercourse.
ARTIFICIAL INTELLIGENCE
Simulated intelligence has been represented as a 'component of software engineering,' which implies that frameworks or methods that analyze and process data in a broad range of uses may be implemented. Including precision-based knowledge is ready to modify medicine, population health and natural language handling; artificial intelligence has been concentrated into a few areas of clinical services. Similar intelligence, for example, radiography, pathology, ophthalmology and dermatology, is therefore deemed relevant to visually oriented claims of renown. The performance of human-made medical intelligence follows two fundamental ideas: virtual and physical. The virtual element in simulated intelligence in medicine is dependent on the AI that works on numerical calculations that enhance comprehension through learning. The second kind of application of framework based information in physical items, clinical devices and sensitively robotic medicine cables (care bots). It is possible and desired for effective information on and the development of instruments and tools to manage diabetes via computerized intelligence. Their usage is also desirable. In order to provide more secure innovation via man-made intelligence, safety plans, welfare savings and procedures are required to be recognized for all possibly specialized frameworks, with all vulnerabilities. Wearable products, mobile phones and other gadgets have been shown to assist monitor and track symptoms and infection of the patient. Specialized improvements are possible. Doctors and medical specialists should allow people to collect information based on computer to assist them manage diabetes in a convincing manner. The following components are components of the P7 concept of social insurance products.
ML IN DIABETES
The connection between people with diabetes, competent human authorities and the social safety system that binds them is shown in diabetes care. Care may be improved from every three viewpoints by man-made understanding. The calculator will create a modified bolus proposal, limited by case-based reasoning, once it is accentuated in physiology of a person with repeated usage. Scientists that build up shut circle pancreas also advance towards ever more multifaceted therapy. Current 'hybrid' frameworks are supported for clinical use and are computerized basal discharge, which requires a manual contribution at meals. Ongoing research using versatile control calculations that have changed based on physiological sources of information such as sugar consumption, speaks for progress towards significantly higher automaticity; however, since producers have, for example, taken 125 million men long insulin siphon readings, AI can encourage the rising and coming years of 'insight.' AI also promotes several apps in a profoundly engaging way to work with customers to assist correct consistency. A model is an application that offers appropriate foods considering the current measurement of glucose and glycemic responses of a person.
The Healthcare Professional
AI talks about a changing perspective of clinical medicine from unbiased described administration methods to informative accuracy. The anticipatory preliminary reason given for example that weight loss intercessions in overweight individuals with type 2 diabetes do not reduce cardiovascular hazards; however, AI-based re-investigation distinguished subgroups of members in which weight reduction was sure to be helpful and found to hid that impact by another subgroup where negative consequences had occurred. When this additional granularity is turned into a clinical direction, over a one-size proposal, it would turn into advice that would benefit both sub-groups.
The Healthcare System
By using the 'system vision,' screening programs that rely on clinical knowledge may be fully robotized. For example, an AI model exhibited contrasting authority levels, with a board of ophthalmologists, in identifying referable diabetic retinopathy. Also encouraging examination is important in the screening of foot ulcers. This approach may lead to fewer people generally safe for referral by experts and higher-risk people benefiting from reduced reference delays. The computations of AI are particularly suitable for identifying those who are at risk by integrating an enormous number of weak indications such as hereditary and metabolic research. If these methods are more effective than the most popular segmental variables for the prediction of diabetes, they may promote intervention and enhance the effectiveness of anti-diabetic initiatives.
The Individual with Diabetes
A simple test to self-organized diabetes speaks to variability in blood glucose level linked to lifestyle variables such as exercise and stress. Smart estimates that are aware of the effect of the on the glucose processing of a person, may overcome this. An 'advanced bolus, for instance.
TECHNIQUES
In the treatment of diabetes, several AI-based methods were used. In addition to certain measures of blood glucose and hemoglobin, diagnosis of diabetes has changed with the development of AI (HbA1c).
Case-based reasoning (CBR)
CBR is an AI method used widely for the treatment of diabetes to address new issues by learning from comparable previous events. The Diabetes Support System is an example of CBR in diabetes treatment. It attempts to identify issues in the blood glucose management automatically, suggest remedies to identified problems and keep in mind the efficient and inefficient treatments for each patient. For different meal circumstances in diabetes, CBR has been utilized to optimize and individualize insulin treatment.
Machine learning and deep learning
Several methods of machine learning were utilized to create digital assistance for care for diabetes. They include supporting vector machines, artificial neural grids, bays naıve, decision trees, random forests, classifying and regressing trees and neighboring k-nearest. Automated screening of blood glucose fluctuation was used in machine learning. Machine learning principle including selection techniques (e.g. random forest, regression of the logistics, reciprocal data,main parameters analysis, variance analysis and the Fisher discriminant ratio), outlier removal techniques, cross-validation protocols and classifying techniques Machines may be used to detect diabetes-prone individuals based on genetic and metabolic variables.
Artificial neural networks
Neural networks were developed to connect, evaluate and generate customized information. Diabetes diagnostics have found specific and large uses for neural network technique. Smart algorithms have been designed to assess the effects on glycemic indices of different variables.
REVIEW OF LITERATURE
Devendra Kumar Mishra, (2020) some of the polls also revealed that the number of patients globally at approximately 600 million would be noted by the end of 2040. Our days have become digitized for the overwhelming bulk of our days. Computerized welfare takes into account the continuous expansion of cooperation between cutting-edge therapeutic advances, development, and computerized correspondence. DM is a disease caused by non-regulated diabetes, which may lead to multi-organ deception in individuals. Due to the advancements in AI and human-made intelligence, which enables early detection and analyzes of DM via a more advantageous automated method than a manual finding.
Giang Thu Vu (2020) the increasing incidence and worldwide burden of diabetes support a greater and more effective management of diabetes and of its complications to prevent, monitor and treat. Artificial intelligence has become more prevalent over the years to complement the diagnosis, treatment and prognosis of the trajectory of diabetes. The goal of this study is to demonstrate a bibliographic analysis in an inclusive scope for the use of artificial intelligence in diabetes. Bibliometric analyzes and latent Dirichlet analysis was coupled in an effort to identify emerging areas of study, artificial intelligence issues and diabetes.
Samer Ellahham (2020) Artificial intelligence (AI) is a rapidly developing area that may change the way it is diagnosed and treated in the face of diabetes – a worldwide epidemic. Machine learning principles have been applied to create algorithms to assist prediction models of diabetes risk or its resulting consequences. Digital therapies have shown that lifestyle treatment has been a well-established procedure for diabetes control. Patients are more empowered to self- manage diabetes, and clinical decision assistance is being provided to both patients and health providers..
Rajiv Singla (2019) Artificial intelligence/machine education (AI/ML) transforms all areas of the medical system of our lives. The use of AI/ML may greatly improve the extent of diabetes management and make it more effective. The enormous burden of diabetes in India is a unique combination of issues and offers a unique opportunity for data availability. The use of this data by all doctors may give India an advantage in this field of study by utilizing electronic medical records. The use of AI/ML would assist us understand and enable us develop customized solutions for our issues.
Hanslal Prajapati et al (2017) FIS suggested, which promotes accuracy and improves efficiency. They constructed and updated a fluid, logical expert diagnostic system. For FIS structure, they utilized dataset-PIMA. In the Matlab fuzzy diabetes diagnostic tool, consisting of eight linguistic inputs as well as 1 DM yield, this Fuzzy-logic control was designed.
Sahoo et al. (2016) are of the view that “huge volume of patient data is created every day from doctor's notes, doctor's prescription, clinical reports, and body sensors. The analysis of healthcare parameters and the prediction of the subsequent future health conditions are still in the infant stage. A Cloud enabled Big Data analytic platform is the best way to analyze the structured and unstructured data generated from healthcare management systems.” “In smart cities, the patients are equipped with variety of electronic devices like mobile networks. The smart devices help to monitor the health status of the patients in hospitals and outdoor environment through Internet.
Lin et al. (2016) are of the view that “variety of huge volume of structured and unstructured patient data are collected through devices equipped with sensors and networks in hospital storage system”. Recently, as described by Tawalbeh et al.(2016) “the advancement in mobile devices such as multi-sensor equipped smart phones is also used as the data collection devices”.
Currently, Yu et al.(2016) have reported that “various data computing techniques have emerged as efficient analytic tools. Many advanced techniques and tools have emerged in the recent past as reported by different authors to analyze healthcare parameters”.
Rallapalli et al.(2016) It is considered that 'Big data analysis is performed using different machine learning techniques to identify the group of patients, illnesses and predictions.' Huang et al.(2015) believe that "available technology currently cannot carry out a large-scale health care analysis and processing. Cloud offers a scalable, parallel and distributed processing environment, and Map Reduce for rapid healthcare data processing as the state-of-the-art technology is presently accessible.
Ibrahim M. Ahmed et al (2015) has developed a method of diabetic diet and implemented for usage in Sudan. This method gives medical guidance and fundamental understanding about diabetic diet to patients. The creation of the suggested Expert System has taken many steps from the issue and needs to be identified, analyzed requirements, acquired knowledge, formalized and designed. Visual prolog has been used to create the graphical user interface and to build the system components. The study was conducted at the military hospital in Khartoum, using hybrid quality research technique. The study aimed primarily at developing a system of medical experts for treating diabetes type 2.
AshwinBelle et al.(2015) "The rapidly expanding area of Big Data analytics has begun to play a key role in the development of healthcare and research methods. It has given instruments for building, managing, analyzing and assimilating huge amounts of organized and unstructured healthcare system information. In the administration of treatment and illness investigation, big data analytics have been used lately. But, because of certain basic issues within the Big Data paradigm, adopt rate and growth of research in this sector is still sluggish." Some of the main difficulties in promising fields of medical research have recently also been addressed.
Lee and Han (2011) Designed and suggested a "window-based interim data collecting and monitoring models for patient monitoring to improve their effectiveness. Patient data are gathered, stored and processed in the conventional medical systems in a traditional way that cannot enable the diagnosis of complicated diseases".
Elshazly et al. (2013) He stated that 'medical imaging, in addition to illness status, provides important information on human body architecture and organ function. The increasing number of patients and health organizations has led to the shared usage of medical diagnostics computers. Decision assistance systems in healthcare contexts are also made more helpful. They found that several elements of the health systems may be enhanced via the use of computer methods".
According to Dougherty (2009) “combining medical images data with other types of data can be able to reduce the time taken for a diagnosis and improve the accuracy of decision-making.”
Dilip Kumar Choubey et al (2014) conducted surveys on a number of soft computing methods, including vector support machinery, colony optimization Antibiotic, neural network, fuzzy logic, genetic algorithm, rude sets, etc. For speeding diabetic prognostics, the investigator created a Web-based expert system. If rules are then used, the web-based application has a knowledge base. The application takes user data from personal information to symptoms. Application Fuzzy system expert capable of categorizing diabetes into type 1, type 2, and pre-diabetes depending on the symptoms entered
Anindito Yoga Pratama et al (2013) the mobile application that is expert system developed android basis. The risk of diabetes is determined by a person. Indeed, it is just a tool for early diabetes prevention. An expert system's primary benefit is its capacity to operate constantly and is subject to maximal condition. The expert system intended not to replace an expert's position, but rather as an instrument of decision making since numerous options might be selected properly.
M. Kalpana et al (2012) the proposed fuzzy expert system offers improved diabetes diagnostic versatility. They analyzed the PIDD data set and converted the crisp value data set to fuse the interface. The correlation determination tool analyzes the impact of the fluorescent operator with rules to decide on the number of members to provide the knowledge descriptions for the practitioner. Finally, the fluid-output converter was adopted by the de-fuzzification interface. Experimental results indicate that this tool can analyze information effectively in comparison with previous instruments.
Danijela TADIC et al (2010) structured new fuzzy model for type 2 patients in particular, to evaluate and to decide for each level the best remedial method. This fuzzy multi-criteria tool decides on several systems that are beneficial for every patient treated. This approach was adjustable according to the likelihood of number change, the kind of progression standards modification and the relevance of improvements.
Abdulla Al-Malaise Al-Ghamdi et al (2011) presented a novel cloud-based T expert system that is web-based, adaptable, simple to use and secure for the treatment of diabetes using Google App Engine. This expert system is accessible to the investigator. This expert cloud computing system was created to address the diabetes issue and to improve management and to accelerate treatment by doctors.
Juan Li (2020) “Application of Artificial Intelligence in Diabetes Education and Management: Present Status and Promising Prospect” Diabetes is an incurable lifelong disease despite the fast advancement of science and technology in healthcare. Diabetes education aimed at improving self-handling is a vital method of enhancing the metabolic control and quality of life of patients. In converting accessible genetic data and clinical information into useful knowledge, artificial intelligence (AI) technology has made tremendous improvements. The use of AI tech in the teaching of diseases is very helpful given their benefits in fostering individualization and full-course intervention based on the unique images of people. In this article the most current uses of AI technology for several areas of diabetes training are reviewed and discussed.
Jyotismita Chaki (2020) “Machine learning and artificial intelligence based Diabetes Mellitus detection and self-management: A systematic review” Diabetes mellitus (DM) is a disease that may progress to multi-organ failure in those with uncontrolled diabetes. Due to advancements in machine learning and artificial intelligence, which make early identification and diagnosis of DM more beneficial than manual diagnosis by means of an automated process? A number of papers on automated DM detection, diagnosis and self-management are now being published using machine learning and artificial intelligence technologies. This study analyzes the approaches used to identify, diagnose and manage the DM in six aspects, i.e. DM datasets, pre-processing methods, extract features, machine learning, classification and DM diagnostics, and artificial intelligence-based intelligent DM assistant and performance measurements.
Mercedes Rigla (2018) “Artificial Intelligence Methodologies and Their Application to Diabetes” The combination of continuous glucose monitoring and insulin pump data revolutionized diabetes treatment in the last 10 years. More recently, a range of functionalities and physiological variables have been accessible via wristbands or watches, for example, heart rate, sleep time, numbers of steps and movement. In the future the new data shall be included, including hydration, geo-location and barometric pressure. If evaluated, all these characteristics may assist patients and physicians to support decisions. In most of the medical disciplines, similar new situations have been developed, so that in recent years more interest has been shown in the development and use of artificial intelligence (AI) techniques for supporting decisions and acquiring information.
RESEARCH METHODOLOGY
Table 1: Research Methodology Adopted
Primary Data Collection in our Research:
DATA ANALYSIS
For these two class problem, researcher define specificity as the number of images correctly identified as diabetic retina (diabretina) and normal retina (normalretina). Researcher defines accuracy in the form of the images of retina with a accurate classification. This final designed network succeeded witth 80% of accuracy. With the use of Fastai libraries it able to categorize any number of image. The trained convolutional neural network makes a rapid prediction a well as immediate reply to user. Following are some of test cases researcher has given. Predict the class of the images for which we created the model. Cloud services offers by google with virtual machines where graphical processing units prepared for more then nine hundred and sixty tera flops, execution for each event. DL, moleculaer modeling, and simulation are animated with V100 and P4, P100, Tesla K80, NVIDIA, T4GPUs. Despite the dimension of your remarkable job needing to be done, anyone can get faultless GPU given by.
Test Cases:
Following are the test cases randomly the researcher has chosen.
Case 1:
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the correct interpretation that is the category is diabetic retina.
Case 2:
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the correct interpretation that is the category is diabetic retina.
Case 3
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the correct interpretation that is the category is diabetic retina.
Case 4:
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the correct interpretation that is the category is diabetic retina.
Case 5:
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the incorrect interpretation that is the category is normal retina where it is diabetic retina.
Case 6:
For this situation researcher used the one of the image from the dataset which is used for testing pourpose for checking class of the image where CNN model given the correct interpretation that is the category is normal retina.
CONCLUSION
The literature survey and current research both point to the many potential uses of artificial intelligence and machine learning in the medical field. Currently, AI and ML are being used to cure infectious illnesses, customize patient care, and streamline hospital administrative procedures. Research in data science shows significant promise for enhancing both the diagnosis of diabetes mellitus and the prediction of diabetes kinds, which would be of great use to both medical professionals and their patients. The process of creating a machine learning based model for diabetes detection is efficient. It is anticipated that the use of AI and ML methods will increase in the field of logic research in the future years. The machine learning model is a cutting-edge approach that blends the strengths of machine learning with those of artificial intelligence. This machine learning model predicts future values of diabetes mellitus and classifies different forms of diabetes using a non-parametric, supervised learning process using a basic decision tree classification method. With the use of simple criteria supplied by the user and learned from the datasets, our machine learning model may predict whether or not a given sign or symptom is indicative of diabetes mellitus and, if so, the kind of diabetes the patient may have. This is a generalized approach that may be used to adapt existing expert systems into a machine learning environment.
REFERENCE
  1. Devendra Kumar Mishra, and Shubham Shukla ROLE OF ARTIFICIAL INTELLIGENCE IN DIABETES MANAGEMENT. International Journal of Engineering Technologies and Management Research, 7(7), 80-88. https://doi.org/10.29121/ijetmr.v7 .i7.2020.728, 2020.
  2. Giang Thu Vu, Bach Xuan Tran, Roger S. McIntyre, Hai Quang Pham, Hai Thanh Phan, Giang Hai Ha Kenneth K. Gwee, Carl A. Latkin, Roger C.M. Ho and Cyrus S.H. Ho Int. J. Environ. Res. Public Health, 17, 1982; doi:10.3390/ijerph17061982, 2020
  3. Samer Ellahham “Artificial Intelligence: The Future for Diabetes Care”, The AmericanJournal of Medicine, Vol 133, No 8, 2020.
  4. Singla, Rajiv & Singla, Ankush & Gupta, Yashdeep & Kalra, Sanjay. Artificial Intelligence/Machine Learning in Diabetes Care. Indian Journal of Endocrinology and Metabolism. 23. 495. 10.4103/ijem.IJEM_228_19.
  5. Hanslal Prajapati1, Anurag Jain2, Sanjay Kumar Pal3 “ An Enhance Expert System for Diagnosis of Diabetes using Fuzzy Rules over PIMA Dataset”, International Journal of Advance Engineering and Research Development (IJAERD) Volume 4, Issue 9, September-2017, e-ISSN: 2348 – 4470.
  6. Sahoo. PK, Suvendu Kumar Mohapatra And Shih-Lin Wu, Analyzing Healthcare BigData With Prediction for Future Health Condition. IEEE Access. Vol. 4, 9786-9799;2016.
  7. Lin,K, F. Xia, W. Wang, D. Tian, and J. Song, "System design for big data application in emotion-aware healthcare'', IEEE Access, vol. 4, pp. 6901- 6909,2016.
  8. Yu, Zhiwen; Luo, Peinan; You, Jane; Wong, Hau-San; Leung, Hareton; Wu, Si;Zhang, Jun; Han, Guoqiang., "Incremental semi-supervised clustering ensemble for high dimensional data clustering'', IEEE Trans. Knowl. Data Eng., vol. 28, no. 3, pp. 701-714, Mar. 2016.
  9. Rallapalli S, R. R. Gondkar, and U. P. K. Ketavarapu, "Impact of processing and analyzing healthcare big data on Cloud computing environment by implementing hadoop cluster,'' Procedia Comput. Sci., vol. 85, pp. 16-22, May 2016.
  10. Ibrahim M. Ahmed, Marco Alfonse, Mostafa ArefDr.Adbel-Badeeh M. Salem, “DailyMeal Planner Expert System for Diabetics Type-2”, E-Leader Prague 2015.
  11. Ashwin Belle, Raghuram Thiagarajan, S. M. Reza Soroushmehr Fatemeh Navidi, Daniel A. Beard, and Kayvan Najarian, Big Data Analytics in Healthcare BioMed Research International Volume 2015.
  12. McAfee et al. have described that “Big Data is a collection of heterogeneous complex data which requires new hardware and software devices to store, analyze and visualize the data successfully. Healthcare data is available in many healthcare centers of researchers, health insurance managements and government organizations. However, these data are not able to deliver comprehensive information to the users”, 2012.
  13. McAfee A. E. Brynjolfsson, T. H. Davenport, D. J. Patil, and D. Barton, “Big data: the management revolution,” Harvard Business Review, vol. 90, no. 10, pp. 60–68, 2012
  14. Lee M and X. Han, "Complex window query support for monitoring streaming data in wireless body area networks'', IEEE Trans. Consum. Electron., vol. 57, no. 4, pp. 1710-1718, Nov. 2011
  15. Elshazly H , A. T. Azar, A. El-korany, and A. E. Hassanien, “Hybrid system for lymphatic diseases diagnosis,” in Proceedings of the International Conference onAdvances in Computing,Communications and Informatics (ICACCI’13), pp. 343–347, IEEE, Mysore, India, August 2013.
  16. Dougherty G, Digital Image Processing forMedical Applications, Cambridge UniversityPress, 2009.
  17. Dilip Kumar Choubey, Sanchita Paul and Joy Bhattachrjee , “Soft Computing Approaches for Diabetes Disease Diagnosis: A Survey”, International Journal of Applied Engineering Research ISSN 0973-4562 Volume 9, Number 21,2014 pp. 11715-11726
  18. Anindito Yoga Pratama, Dewi Agushinta R., and Remi Senjaya ,“Design Of Mobile Expert System For Diabetes Risk Diagnosis And Information”, Journal of Information Systems, Volume 9, Issue 1, April 2013, 32-36
  19. M. Kalpana, Dr. A.V Senthil Kumar,“Diagnosis of Diabetes using Correlation fuzzy logic in Fuzzy Expert System”, International Journal of Advanced Research in Computer Science, 3 (1), Jan –Feb, 2012,244-250
  20. D., Tadić, P., Popović, A., Đukić,“A Fuzzy Approach to Evaluation” Yugoslav Journal ofOperations Research Volume 20, 2010, Number 1, 99-116 , 10.2298/YJOR1001099
  21. Abdulla Al-Malaise Al-Ghamdi et al, “An Expert System of Determining Diabetes Treatment Based on Cloud Computing Platforms”, (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 2 (5) , 2011, 1982-1987
  22. Li, J., Huang, J., Zheng, L., & Li, X. Application of Artificial Intelligence in DiabetesEducation and Management: Present Status and Promising Prospect. Frontiers in public health, 8, 173. https://doi.org/10.3389/fpubh.2020.00173, 2020.
  23. Jyotismita Chaki, S. Thillai Ganesh, S.K Cidham, S. Ananda Theertan, “Machine learning and artificial intelligence based Diabetes Mellitus detection and self- management: A systematic review”, Journal of King Saud University - Computer and Information Sciences, 2020, ISSN 1319-1578,https://doi.org/10.1016/j.jksuci.2020.06.013.
  24. Mercedes Rigla “Artificial Intelligence Methodologies and Their Application toDiabetes”, Journal of Diabetes Science and Technology 2018, Vol. 12(2) 303–310