Fuzzy inference systems for predicting and evaluating diabetes likelihood and severity

Introduction

Diabetes mellitus, a complicated metabolic issue characterized by increased levels of glucose in the blood is an enormous and emerging health care issue in the world. The varied concept of diabetes demands extensive and accurate symptomatic and prognostic equipment to guide clinicians evaluate its severity and initiate potent treatment regimens. Recently, foreseeing and evaluating the gravity of diabetes via Fuzzy inference system (FIS) has emerged as a promising direction to follow, to provide a subtle methodology to address the inherent weaknesses and inaccuracies of clinical data. Fuzzy rationale is excited by human dynamic cycles which often involve obscure or equivocal data that provide a framework of showing the complicated relationships using phonetic factors and the principle of fuzziness. FIS is especially well-adapted to the use in the area of healthcare since it can be applied to develop intelligent systems, which could capture the complexities and peculiarities inherent in the medical diagnosis. The commonly applied demonstrative methods of surveying diabetes severity often rely on the limit based methodology that might fail to adequately encompass the variability of the consistent information and the interaction of a vast number of influencing variables. FIS, in its turn, manages to address the uncertainties, and this is why it is a powerful tool to decipher the multifaceted relationships between clinical limits such as blood glucose levels, HbA1c values, and weight file (BMI) and the severity of diabetes.

Mansourypoor and Asadi (2017) [6] developed a determination framework of diabetes mellitus using a support learning based transformative fuzzy rule-based framework. Birjais et al. (2019) [5] have tried to focus more on the diagnosis of diabetes that is one of the chronic diseases increasing at an alarming rate globally as per world health organization in 2014. Naz and Ahuja (2020) [7] examined the different machine learning methods of classification in a bid to predict diabetes. ANN, NB, DT and deep learning (DL) were compared. Pradhan et al. (It was proposed to detect diabetic patients with a model based on artificial neural networks (ANNs) in 2020) to detect them. The functioning of the proposed model was tested with the assistance of Pima Indian Diabetes (PID) dataset. The article by Tigga and Grag (2020) [10] developed an algorithm to categorize type 2 diabetes through machine learning. The model was oriented to the achievement of the gamble of diabetes in the minds of individuals in regards to their lifestyle and family background. In order to spearhead the study, an on the web and disconnected based poll was conducted to obtain 952 instances of diabetes. The Akpado et al. (2021) [3] developed a model of a fuzzy logic-driven expert system that would be used in diabetes type 2 diagnosis. The developed expert system contained the four steps of Mamdani fuzzy inference system namely fuzzification, rule evaluation, output aggregation and difuzzification. Arora et al. (2021) [4] suggested the use of fuzzy-based system to predict Covid-19 is rooted in the symptoms and the parameters of an individual. It receives input limits such as fever, hack, breathing difficulty, muscle sore, sore throat, traveling history, age and clinical history in the form of different enrollments works and generates one output which estimates the possibility of an individual being infected with coronavirus using Mamdani fuzzy inference framework. Aamir et al. (2021) [1] applied fuzzy rationale to foster an interpretable model and to play out an early determination of diabetes. Fuzzy rationale has been joined with the cosine plentifulness strategy and two fuzzy classifiers have been built. Thakkar et al. (2021) [9] proposed a method that provides a pipeline of various tasks, such as selecting the dataset and preprocessing the data using standardization, normalization, and other techniques. Aggarwal et al. (2022) [2] proposed a coronavirus risk expectation model for diabetic patients utilizing a fuzzy surmising framework and AI draws near. Their study sought to estimate the level of COVID-19 risk in diabetic patients who did not receive prompt medical advice and overcome the multifold COVID-19 mortality rate in diabetic patients. In this review study, Yazdani et al. Machine learning (ML) methods were developed and validated using the data of 784 elderly people in 2023 [11] . In this paper, we dig into the development and approval of a FIS model for diabetes seriousness expectation. We examine the choice of significant clinical boundaries, the plan of fuzzy guidelines, and the foundation of participation capabilities to catch the innate vulnerability in clinical information. In addition, we investigate the potential of FIS for assessing the impact of various treatment strategies on diabetes severity, contributing to ongoing efforts to improve diabetes care's precision and individualization.

Membership Plot Functions of Input and Output Variables

By defining how input values relate to fuzzy sets and, consequently, how these fuzzy sets contribute to the overall fuzzy logic model, membership functions play a crucial role in Fuzzy Inference Systems (FIS). A participation plot outwardly addresses these capabilities and gives bits of knowledge into the fuzzy rationale framework's way of behaving.

Rule Base

The fundamental elements that establish the relationships between the input variables, fuzzy sets, and the output variable are the rules of a fuzzy inference system (FIS) for predicting and evaluating the severity of diabetes. The system's decision-making logic is defined by these rules, which direct how input data is processed to make accurate predictions.

Defuzzification 4(i). Likelihood of Diabetes

The centroid can be calculated by

Estimated Values of Likelihood and Severity of DIABETES for Different Input Parameters

The assessed upsides of probability and seriousness of diabetes for various information boundaries in a fuzzy inference system (FIS) give significant bits of knowledge into the potential results in light of the fuzzy rationale model.

3D Surface Visualization of Fuzzy Inference Output for DIABETES Risk

A 3D surface perception of fuzzy induction yield for diabetes risk gives a strong means to investigate and figure out the intricate connections between input factors, fuzzy principles, and the subsequent gamble evaluations. The graphs ( 1) and ( 2) indicate that the points are the combinations of the BMI and the level of blood glucose. The surface height at the point is a measure of the possibility and the severity of diabetes respectively. Places of the drawing where the area is elevated or more opaque can evince an increased gamble of diabetes. Risk, on the contrary, is reduced in lower areas. The patterns, trends, or clusters of the graphs illuminate the interaction between the body mass index (BMI) and the blood glucose level in the prediction of the risk and intensity of diabetes. The graphs (3) and ( 4) can be unraveled using the association of different combinations of blood glucose level and circulatory strain with the likelihood of diabetes and its severity. This could help illuminate trends and possible risk factors related to the probability and the intensity of diabetes. In the figures ( 5) and ( 6), data points correspond to a person or group of people sharing specific values of the degree of blood glucose, age, and probability/severe of one having it. Aspects that are critical are those where both the probability and the extent is high. Patterns, clusters, or trends which provide light on such insights as what age groups are more prone to develop diabetes or what blood glucose level is correlated with an increased risk. Surveillance plot on diagrams ( 7) and ( 8) reveal the effect of the combination of BMI and pulse on the probability or severity of diabetes. Every point on the superficial level deals with a specific combination of a BMI and a pulse and the level of a surface by that point indicates the value of likelihood or severity. As we travel along the X-hub (varying BMI esteems), the graphs indicate the variation of probability or seriousness. Equally, the charts illustrate the effect on likelihood or severity that would occur as we traverse the Y axis (varying blood pressure values).

Two surface plots, one of the likelihood of diabetes and the other of the severity have been plotted in graphs ( 9) and (10), respectively. This involves matching the fragments of information with surfaces to demonstrate how the factors alter with the different mixes with old enough age and circulatory strain. When we observe these 3D diagrams, we observe examples and trends in the probability as well as the severity of diabetes. It is observed that with increasing age, the likelihood of diabetes will in most cases increase and the severity can also increase.

Performance Metrics of the Model

The estimations of F1 score, exactness, particularity (true negative rate), and accuracy (positive predictive value) from a disarray framework for a fuzzy inference system (FIS) anticipating and assessing diabetes seriousness are as per the following:

Conclusion

Overall, the introduction of fuzzy inference system (FIS) to predict and detect the seriousness in diabetes is one major step to overcoming the limitations of the traditional symptomatic approach. The natural complexity of diabetes, which is illustrated by its multi factorial complexity and varying degrees of severity in individuals, demands innovative approaches which are able to capably investigate the intricacies of clinical data. By committing the vulnerabilities and the intrinsic inaccuracy in the clinical data, FIS is a promising path towards achieving this accuracy since it is built on the fuzzy rationale standards. FIS makes a unique range of features visible in the situation of defining the severity of diabetes as discussed earlier. FIS can be used to obtain the nuances of patient data through the use of linguistic variables, fuzzy rules, and well-crafted membership functions. Consequently, it is able to give a more context sensitive and finer analysis of the severity of diabetes. Such flexibility is an emergency in an area where individual types and progressive shifts in wellbeing status play crucial roles in illness circulation.

The FIS ability is able to extend further beyond being anticipatory of seriousness; It is capable of determining the efficacy of different treatment modalities. FIS is a valuable resource to medical practitioners as it can receive dynamic inputs and adapt to evolving conditions of patients. It is this flexibility that ensures that the FIS model can progress at the side of the patient and help in enhancing personalized and receptive treatment strategies. Such introduction is a prelude to a deeper exploration of the subject matter because it presents a general review of the principles on which FIS is based and how it is applied to predicting the severity of diabetes. The lessons learnt in this line can greatly enhance the quality and customized approach to managing diabetes as researchers keep on refining and verifying FIS models on the basis of actual clinical information. The inclusion of FIS into the process of assessing the severity of diabetes, in its turn, is the testimony to the shifting nature of healthcare technology. It clears pathways to more precise, multifaceted, and tailor-made intercessions, to the general aim of attending to ongoing outcomes and individual contentment to the affected individuals by diabetes. The anticipated impact of FIS in altering the healthcare of diabetes continues to emerge as the examination moves forward, providing a ray of hope in the ongoing effort to manage the global epidemic of diabetes.

References

1. Aamir KM, Sarfraz L, Ramjan M, Bilal M, Shafi J, Attique M, et al. A fuzzy rule-based system for classification of diabetes. Sensors. 2021;1-17.
2. Aggarwal A, Chakrador M, Bhatia MS, Kumar M, Stephan T, Gupta SK, et al. COVID-19 risk prediction for diabetes patients using fuzzy inference system and machine learning approaches. Journal of Healthcare Engineering. 2022;4096950:1-10.
3. Akpado KA, Njonu UJ, Obioma PC, Isizoh AN. An improved method for predicting diabetes mellitus using adaptive neuro-fuzzy inference system. International Journal of Scientific Engineering and Science. 2021;5(11):19-32.
4. Arora S, Vadhera R, Chugh B. A decision-making system for corona prognosis using fuzzy inference system. Journal of Fuzzy Extension and Application. 2021;2(4):344-54.
5. Birjais R, Mourya AK, Chauhan R, Kaur H. Prediction and diagnosis of future diabetes risk: A machine learning approach. SN Applied Sciences. 2019;1-8.
6. Mansourypoor F, Asadi S. Development of a reinforcement learning-based evolutionary fuzzy rule-based system for diabetes diagnosis. Computers in Biology and Medicine. 2017;91:337-52.
7. Naz H, Ahuja S. Deep learning approach for diabetes prediction using PIMA Indian dataset. Journal of Diabetes and Metabolic Disorders. 2020;19:391-403.
8. Pradhan N, Rani G, Dhaka VS, Poonia RC. Diabetes prediction using artificial neural network. In: Deep Learning Techniques for Biomedical and Health Informatics. 2020;327-39.
9. Thakkar H, Shah V, Yagnik H, Shah M. Comparative anatomization of data mining and fuzzy logic techniques used in diabetes prognosis. Clinical eHealth. 2021;4:12-23.
10. Tigga NP, Garg S. Prediction of type 2 diabetes using machine learning classification methods. International Conference on Computational Intelligence and Data Science. 2020;167:706-16.
11. Yazdani A, Shanbehzadeh M, Kazemi-Arpanahi H. Using an adaptive network-based fuzzy inference system for prediction of successful aging: a comparison with common machine learning algorithms. BMC Medical Informatics and Decision Making. 2023;23(229):1-14.