CC BY
4IMPORTANCE OF FUZZY LOGIC METHODS IN SOLVING PROBLEMS OF MEDICAL DIAGNOSIS AND PROGNOSIS 1Khidirova Charos Murodilloyevna, Jabborova Nozima Sattor kizi
1Tashkent University of Information Technologies named after Muhammad al-Khwarizmi. Associate Professor of the Department of Software of Information Technology, 2Ph.D. student at Tashkent University of Information Technologies named after Muhammad al-Khwarizmi. 1khidirova@tuit.iuz, 2jabbarovan1006@gmail.com https://doi.org/10.5281/zenodo.1072129 7
Abstract. The article talks about how diagnosing patients forms an important part of medical practice and medical science, as well as the digitization of this process. In the process of making this diagnosis, experts in the field perform it intuitively based on their experience. At the same time, as information technologies are rapidly penetrating all fields, they are also being integrated into the medical field. Digitization of knowledge and individual experience of medical specialists is one of the urgent problems. Even though the computerization of the medical diagnosis process began in the 50s of the last centuries, some issues are still waiting to be solved. Since such problems are a problem of complex formation, the use of soft calculation methods of computer modeling in solving such problems allows to obtain more accurate solutions. The article provides analytical information about the use of fuzzy logic methods in making decisions in the medical diagnosis of patients.
Keywords: Chronic diseases, fuzzy logic, edge computing, neuro-fuzzy, prediction, diagnosis, dataset, diagnosis, fuzzification, defuzzification, black box, symptoms, factors, clustering.
Аннотация. В статье рассказывается о том, как диагностика пациентов является важной частью медицинской практики и медицинской науки, а также о цифровизации этого процесса. В процессе постановки этого диагноза эксперты на местах проводят его интуитивно на основе своего опыта. В то же время, поскольку информационные технологии быстро проникают во все области, они также интегрируются в медицинскую сферу. Одной из актуальных проблем является оцифровка знаний и личного опыта медицинских специалистов. Несмотря на то, что компьютеризация процесса медицинской диагностики началась в 50-е годы прошлого века, некоторые проблемы все еще ждут своего решения. Поскольку такие задачи являются проблемой комплексного формирования, использование мягких методов расчета компьютерного моделирования в решении таких задач позволяет получить более точные решения. В статье представлена аналитическая информация об использовании методов нечеткой логики при принятии решений при медицинской диагностике пациентов.
Ключевые слова: хронические заболевания, нечеткая логика, периферийные вычисления, нейро-нечеткий прогноз, диагноз, набор данных, диагноз, фаззификация, дефаззификация, черный ящик, симптомы, факторы, кластеризация.
Annotatsiya. Maqolada bemorlarga tashxis qo'yish tibbiy amaliyoti va tibbiyot fanining muhim qismini hosil qilishi hamda bu jarayonni raqamlashtirish haqida so'z yuritiladi. Ushbu tashxis qo'yishjarayonida soha mutaxasislari tomonidan o'z tajribalaridan kelib chiqib intuitive amalga oshiradilar. Shu bilan birga axborot texnologiyalari barcha sohalarga jadal suratda kirib borgani kabi, tibbiyot sohasiga xam integratsiya bo'lmoqda. Tibbiyot soha mutaxassislarini bilim va individual tajribalarini raqamlashtirish dolzarb muammolardan biridir. Tibbiy tashxis qo'yish jarayonini komputerlashtirish o'tgan asrni 50 yillaridan boshlanganiga qaramasdan xozirgi
vaqtgacha echimini kutayotgan masalalar mavjud. Bunday masalalar murakkab shakllanuvchi masala bo'lganligi sababli, bu kabi masalalarni yechishda computer modellashtirishning yumshoq hisoblash usullaridan foydalanish bir muncha aniqroq echimlarni olishga imkon beradi. Maqolada bemorlarni tibbiy tashxislashda qaror qabul qilishda noravshan mantiq usullaridan foydalanish holatlari xaqida taxliliy ma'lumot keltirilgan.
Kalit so'zlar: Surunkali kasalliklar, noravshan mantiq, "edge computing", neyro-fuzzy, bashoratqilish, tashxis qo'yish, ma'lumotlar to'plami, diagnostika, fuzzifikatsiya, defuzzifikatsiya, qora quti, symptom, omil, klasterlash.
I. Introduction
The analysis of scientific and practical resources in the international network reveals that the use of digital diagnostic systems has become a pressing issue in modern medicine. These digitized diagnostic systems enable rapid analysis of patient information and generate reports that aid in decision-making. These reports help specialists determine the next steps in monitoring the patient's condition and predicting future conditions. Moreover, storing all the necessary information about the patient in a database allows specialists to make accurate diagnoses at any time. This results in quick, precise, and effective diagnosis and treatment of the patient's condition. Digitized diagnostic systems have been used in medicine since the second half of the last century and are continuously being improved. Today, experts in the field are working together to take digitized medical diagnosis systems to a new level by widely applying artificial intelligence methods and algorithms to medical diagnosis. The continuous improvement of machine learning, deep learning, and intelligent analysis of meta-data within the framework of artificial intelligence makes it possible to develop intelligent systems for medical diagnosis that provide fast, accurate, and reliable information. The variety of symptoms that serve to determine the patient's condition in medical diagnosis and the lack of a universal metric to measure them show that it is a complex and difficult problem. The lack of a universal merican suitable for different categories of symptoms can lead to many ambiguities.
Based on the source and literature review, fuzzy logic is an emerging field that can create hybrid systems for disease detection and prediction. This logic is the closest approximation to the appropriate level of probability based on true or false outcomes. In this article, the term "unclear" does not mean vague but reflects the incomplete knowledge that exists in the medical field. Fuzzy logic can stimulate the interpretation phase in disease diagnosis, aiding in diagnosis. Therefore, the main goal of this article is to determine its trend and effectiveness in disease diagnosis through a systematic analysis of studies using fuzzy methods related to various medical fields and diseases [2].
II. Main part
Medical diagnosis is crucial for the control and management of infectious diseases. Doctors form a set of symptoms of the disease with the help of necessary sources and diagnose the patient through these symptoms. Two aspects are important in developing a computer model of the diagnosis process: assessing the patient's condition using the symptoms of the disease and making a diagnosis based on the resulting assessments. Computational tools are essential for understanding epidemiological patterns in the spread of disease [3].
Since 2017, research on the broad interpretation of Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) models has grown significantly. This trend indicates a
global demand for transparent and interpretable AI models in various fields, including disease diagnosis.
- interpretability,
explainability,
nable AI
Figure 1. Google Trends data indicates a rising interest in interpretability, explainability, and explainable AI worldwide since 2017.
The theory of fuzzy sets, introduced by Zadeh [1], is used to deal with uncertain data. Fuzzy logic provides membership levels for object properties and uses membership functions to describe ambiguity. There are different types of membership functions like triangle, trapezoid, and Gaussian. However, fuzzy logic is not capable of automatically learning models from data [4]. Neural networks have solved this problem, but they are not suitable for modeling when the data is incomplete or uncertain. To overcome these issues, the neuro-fuzzy model, which is a hybrid model of neural networks and fuzzy logic, was introduced. The adaptive neuro-fuzzy inference system (ANFIS) [5] is one such method that can automatically learn models from uncertain data (as shown in Figure 2). The proposed method combines fuzzy logic, linguistic reasoning, and dimensionality reduction to solve the problems faced by multi-criteria recommendation systems. There are different approaches to combining the predictors' results, including mean prediction, majority voting, and nonlinear decision functions. However, there is no conclusive evidence to suggest that more complex combination approaches work better. The study considered three types of HF: triangular, trapezoidal, and Gaussian. Defuzzification is performed using the Centroid of Area (COA) method, which is one of the most widely used defuzzification strategies. This method calculates the centroid of the fuzzy set A by computing the centroid of the area under the membership function A(z). Using the COA method, the exact value of the fuzzy set A can be calculated.
Zcoa —
f /uA(z)zdz
Jz VA(z)zdz
Diagnosing medical diseases is a challenging and complex task for doctors. They must consider various factors, circumstances, and medical evidence to diagnose patients. However, disease diagnosis can be prone to errors due to its uncertain and intricate nature. This can lead to significant uncertainty because different patients may have different levels of specificity for different diseases. Moreover, computer-aided diagnosis models must have a clear thought process to consider all factors. A decision made by a transparent and understandable model, using the same information as a black box, can be more reliable for doctors. This helps reduce the added
uncertainty introduced by the computer-aided diagnostic model and the difficulty in recognizing hidden relationships between factors. Figure 2 illustrates a computer-aided diagnosis system where an interpretable model produces a decision that is easier for physicians to understand than a blackbox model [6].
Data preprocessing functions, data collected by specialist doctors, i.e. patient data x = (xi, X2, X3,...., xn) given above, collection and processing of disease symptoms, int, real., or Boolean, because of processing it is taken as x = (xi, X2, X3,...., Xn). In the figure below, we can compare the digitalization of the task of prediction and diagnosis, i.e. diagnosis by a computer program, and the diagnosis performed by a specialist.
Fig. 2. Illustration of a computer-aided diagnosis system. Compared with a decision made using a black box model, the decision made using an interpretable model is more understandable for doctors.
Medical decision-making involves two approaches: differential diagnosis (DD) and provisional diagnosis (PD). DD involves taking the patient's medical history and characteristics as input, followed by security knowledge-based analysis to identify and investigate overlaps with similar symptoms. Optional weights are assigned to each diagnosis to represent the overall burden, and these weights are evaluated through multiple clinics and new ones [7].
Sometimes, different diseases can present with similar symptoms, making it difficult to make an accurate diagnosis. To address this, doctors attempt to match the classical condition of each possible DD by measuring similar symptoms. They then rank possible diagnoses and suggest treatment strategies.
The best-matched disease risk is the final diagnosis (PD) after examinations and initial treatment. This diagnosis is closely related to all possible outcomes from which the PD is obtained. In this study, fourteen symptoms of depression were examined according to the index and DSMIV-TR code. These symptoms represent independence, while corresponding levels of depression indicate a dependent factor. A panel of five senior psychiatrists with an average of 10.4 years of experience participated in the study. Psychiatrists assigned weights between 0 and 1 to each symptom/factor and level of depression (mild, moderate, and severe). The severity of symptoms reflects the severity of primary care. Based on the clinical care and knowledge base of psychiatrists, "symptom quality" relationships are possible. The data set has an internal worker
and reliable PCA with a calculated Cronbach's a of 0.87, which exceeds the cutoff of 0.7. PCA measures the eigenvectors and eigenvalues of the correlation matrix, with higher eigenvalues representing principal components (PCs) under study. In this case, only latent (availability, power, "S"), "P" (anhedonia), "W1" (weight loss), "I1" (insomnia) with eigenvalues greater than 1.0, "H" (Hypersomnia), "A" (loss of appetite) and "P1" (psychomotor agitation), which make the main components (PC). This breaks down the original 14-dimensional vector space into 7 dimensions, like how psychiatrists extract important symptoms from a larger set during differential diagnosis. Finally, after rescaling the data set to 302 x 7, a is checked and found to be 0.72, which is above the threshold of 0.7 [8].
Breast cancer is a significant public health issue that causes the highest mortality rate among women around the world. Fortunately, early detection, screening techniques, and awareness campaigns have helped to reduce breast cancer deaths. This paper discusses a methodological approach using Cellular Automata (CA) to identify and segment suspicious regions in mammographic images. CAs are spatially accurate dynamical systems that follow local transition rules, making them easy to understand and intuitive as compared to more complex methods, such as neural networks. The proposed CA-based technique provides an efficient way to detect anomalies within mammograms based on local texture and intensity similarity. The research aims to present an improved version of a cellular automated approach to detect suspicious regions in mammograms. The paper contextualizes this approach within the existing literature, acknowledging alternative methodologies, and justifying the preference for Cellular Automata due to their simplicity and adaptability to the problem [9]. The proposed methodology involves a detailed step-by-step process, including important steps such as image preprocessing, region of interest (ROI) definition, seed selection, and application of a cellular automaton for segmentation. The text explains the importance of each preprocessing step, such as noise reduction, background removal, and pectoral muscle removal, in the context of eliminating potential errors in mammogram analysis. The inclusion of Fuzzy Logic in the cellular automaton-based segmentation process is also explained, which helps to clarify boundaries and increase the algorithm's flexibility. Gray distribution analyses and fuzzy membership functions are introduced as elements that improve the ability of the algorithm to soften the edges of the region of interest. The introduction of fuzzy logic is considered a new contribution to improving the accuracy and flexibility of the segmentation process in the context of medical image analysis [10].
Conclusion
In summary, it is recommended to use fuzzy logic methods in medical and infectious disease diagnostics, interpretive artificial intelligence, machine learning, and visualization for data analysis in solving complex problems related to diseases such as hepatitis, neurological diseases, breast cancer, and infectious diseases. Fuzzy logic methods play an essential role in detection in various medical practices, as observed in our comprehensive review of the latest sources for some diseases. Our analysis of mammography screening for breast cancer assessment shows that fuzzy logic can improve and provide clear contours loaded from cellular automata, resulting in a customized and improved application for medical use. Due to the high expression power and activity of fuzzy logic, it is recommended to adopt fuzzy services for medical information, especially in medical diagnosis. The theory of fuzzy logic in artificial intelligence for medical diagnosis is also emphasized. As the influence of fuzzy logic in artificial intelligence for medicine increases, the conducted analytical research will significantly contribute to the development of medical diagnosis and prediction issues.
REFERENCES
1. L.A.Zadeh. "Fuzzy Sets" https://doi .org/10.1016/S0019-9958(65)90241 -X
2. C.Kupper. "Dietary Guidelines and implementation for celiac disease". https://doi.org/10.10 53/j.gastro.2005.02.024
3. Goli Arji, Hossein Ahmadi, Mehrbakhsh Nilashi, Tarik A. Rashid, Omed Hassan Ahmedg, N ahla Aljojo, Azida Zaino. "Fuzzy logic approach for infectious disease diagnosis: A methodi cal evaluation, literature, and classification". Biocybernetics and Biomedical Engineering Vo lume 39, Issue 4. https://doi.org/10.1016/j.bbe.2019.09.004
4. Soto J, Melin P, Castillo O. (2013, April). A new approach for time series prediction using e nsembles of ANFIS models with interval type-2 and type-1 fuzzy integrators. In Computatio nal intelligence for financial engineering & economics (CIFEr), 2013 IEEE Conference on (p p. 68-73). IEEE.
5. Jang JSR. ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans Syst Man Cy bern 1993;23(3):665-85. http://refhub.elsevier.com/S 1876-0341(18)30149-7/sbref0170
6. Nilashi M, bin Ibrahim O, Ithnin N. "Multi-criteria collaborative filtering with high accuracy using higher order singular value decomposition and NeuroFuzzy system". Knowl Based Sys t 2014; 60:82-101. https://doi.org/10.1016/j.knosys.2014.01.006
7. Jin CAO, Ta Zhou, Shaohua Zhi, Saikit Lam, Ge REN, Yuanpeng ZHANG, Yongqiang Wan g, Yanjing Dong, Jing Cai. "Fuzzy Inference System with Interpretable Fuzzy Rules: Advan cing Explainable Artificial Intelligence for Disease Diagnosis-A Comprehensive Review". ht tps://doi.org/10.1016/j.ins.2024.120212
8. Subhagata Chattopadhyay. "A neuro-fuzzy approach for the diagnosis of depression". https:/ /doi.org/10.1016/j.aci.2014.01.001
9. Iulia-Andreea Ion, Cristiana Moroz-Dubenco, Anca Andreica. "Breast Cancer Images Segme ntation using Fuzzy Cellular Automaton". Procedia Computer Science Volume 225, 2023, P ages 999-1008. https://doi.org/10.1016/j.procs.2023.10.087
10. Qinting Jiang, Xuanhong Zhou, Ruili Wang, Weiping Ding, Yi Chu, Sizhe Tang, Xiaoyun Ji a, Xiaolong Xu. "Intelligent monitoring for infectious diseases with fuzzy systems and edge computing: A survey". https://doi.org/10.1016/j.asoc.2022.108835
