Fuzzy Logic Approaches in Breast Cancer Risk Prediction: A Review

Review Article Correspondence to: Tahir Abdulhakim, e-mail: [email protected] Copyright: © 2025 The authors. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License. How to Cite: Abdulhakim et al. (2025). Fuzzy Logic Approaches in Breast Cancer Risk Prediction: A Review. Scholar J Computational Science, 2(11). DOI: 10.5281/zenodo.17505928 Fuzzy Logic Approaches in Breast Cancer Risk Prediction: A Review Tahir Abdulhakim, Usman Umar Faruk, Jibril Muhammad and Bako Halilu Egga Department of Computer Science, Nasarawa State University, Keffi, Nigeria. Received 18 September 2025; Acceptance 13 October 2025; Published 2 November 2025. Abstract Breast cancer remains a leading cause of cancer-related deaths among women worldwide. Conventional risk prediction models often struggle with the complexity and uncertainty inherent in breast cancer risk factors. Fuzzy logic, introduced by Zadeh in 1965, offers a mathematical framework to handle imprecision and linguistic reasoning, making it a promising tool for early breast cancer risk assessment. This review summarizes key research developments on fuzzy logic–based models for breast cancer risk prediction, explores their methodologies, evaluates their strengths and limitations, and highlights future research needs. Findings indicate that integrating fuzzy inference systems with user-friendly web or mobile platforms improves accessibility and personalized risk assessment, potentially supporting early intervention strategies and reducing breast cancer burden. Keywords: Breast cancer, Fuzzy logic, Risk prediction, Expert systems, Health informatics Introduction Breast cancer is the most frequently diagnosed cancer and the leading cause of cancer-related deaths among women worldwide, accounting for 2.3 million new cases and approximately 685,000 deaths in 2020 alone [1]. Its global burden continues to rise, particularly in lowand middle-income countries where limited access to screening and diagnostic services often delays detection [2]. Despite advances in therapeutic strategies, survival rates remain strongly linked to the stage at diagnosis, underscoring the urgent need for reliable early risk prediction tools [3], [4]. Conventional breast cancer risk models—such as the Gail Model, Tyrer-Cuzick Model, and other regression-based approaches—primarily rely on demographic, hormonal, and familial factors to estimate an individual’s probability of developing the disease [5], [6]. While these models have been widely used in clinical and population-level screening, they exhibit significant limitations. Most employ crisp threshold- Scholar J Computational Science 91 based classifications that fail to capture the inherent uncertainty and imprecision of clinical data, such as “early menarche” or “moderate alcohol intake” [7], [8]. Additionally, they often neglect nonlinear interactions among multiple risk factors, leading to reduced predictive accuracy in heterogeneous populations [9]. To address these challenges, researchers have turned to soft computing techniques, particularly fuzzy logic, introduced by Zadeh in 1965 as a mathematical framework capable of representing partial truths [10]. Fuzzy logic departs from the binary true/false paradigm of classical logic by allowing variables to have degrees of membership between 0 and 1, enabling models to mimic the way clinicians often reason with qualitative descriptors (e.g., “slightly overweight,” “very high risk”) [11], [12]. Mamdani and Assilian pioneered the use of fuzzy inference systems (FIS) to translate expert knowledge into rule-based decision models—an approach now widely adopted in engineering and healthcare decision support [13]. In the context of breast cancer risk assessment, fuzzy logic has proven valuable for incorporating linguistic variables such as age at first pregnancy, age at menarche, duration of breastfeeding, body mass index (BMI), alcohol intake, and smoking status [14]–[16]. By mapping these inputs to triangular, trapezoidal, or Gaussian membership functions and applying IF–THEN rules, fuzzy models offer flexible, interpretable, and human-like reasoning for estimating individual risk levels. Several studies have demonstrated the potential of fuzzy-based systems in breast cancer prediction. Latha developed a clinical decision-support model that processed tumor-related risk variables using fuzzy rules [17]. Yilmaz and Ayan proposed a multi-level fuzzy inference model that achieved 78.3% accuracy in predicting risk categories [15]. Awodele et al. created a mobile-based fuzzy expert system for preliminary risk assessment [16], while Gupta et al. integrated fuzzy logic with decision trees to enhance diagnostic precision [18]. These studies collectively highlight the adaptability of fuzzy systems, although challenges such as limited rule-bases, platform restrictions, and lack of large-scale validation remain [19], [20]. Given these developments, there is growing recognition that fuzzy inference systems can complement conventional statistical models and machine-learning approaches by improving early risk prediction and accessibility, particularly in resource-constrained settings. This review synthesizes current research on fuzzy logic-based models for breast cancer risk assessment, highlighting their methodological foundations, reported performance, and future research opportunities. Literature Search Approach Relevant literature on fuzzy-logic–based models for breast-cancer risk prediction was identified through a focused search of PubMed, Scopus, IEEE Xplore, and Google Scholar using combinations of the terms “fuzzy logic,” “fuzzy inference system,” “breast cancer,” and “risk assessment.” Priority was given to peerreviewed articles published between 2000 and 2024 that described fuzzy inference systems or hybrid fuzzy–machine-learning approaches applied to breast-cancer risk stratification or decision support. Additional references were obtained by screening citations of key papers. Studies that dealt solely with Scholar J Computational Science 92 tumor imaging or histopathology without a risk-prediction component were excluded. This narrative review emphasizes conceptual development, model architecture, and reported clinical relevance rather than quantitative meta-analysis. Evolution of Fuzzy Logic in Healthcare Foundations of Fuzzy Logic The concept of fuzzy logic was first introduced by Lotfi A. Zadeh in 1965 as an extension of classical Boolean logic to handle uncertainty, imprecision, and partial truths [10]. Unlike binary logic, which constrains variables to two discrete states (0 or 1), fuzzy logic allows any value between 0 and 1 to represent the degree of membership of an element in a set. This framework mirrors the way humans interpret qualitative concepts such as “slightly overweight” or “very high risk,” which cannot be defined by strict thresholds [11]. Fuzzy logic evolved rapidly in the 1970s–1980s with the introduction of fuzzy inference systems (FIS) and the Mamdani–Assilian model for rule-based reasoning [13]. FIS uses a set of linguistic IF–THEN rules to link input variables to outcomes, making it particularly suitable for problems with complex, nonlinear relationships. Early adoption was most notable in industrial process control [21] but later spread into domains that required decision-making in the face of vague or incomplete information. Rise of Soft Computing in Medicine Healthcare is inherently characterized by ambiguous data, subjective assessments, and interdependent risk factors. Traditional statistical models often fail to fully capture these complexities, leading to the exploration of soft computing techniques such as fuzzy logic, neural networks, and genetic algorithms [12], [22]. One of the earliest medical applications of fuzzy systems involved clinical decision support systems (CDSS), where expert knowledge was encoded as fuzzy rules for tasks such as diagnosis, treatment selection, and patient triage [23]. Fuzzy logic proved particularly effective in:  Interpreting imprecise laboratory data (e.g., borderline test values)  Modeling qualitative clinical criteria (e.g., “moderate fever,” “mild pain”)  Supporting diagnostic reasoning in oncology, endocrinology, and infectious disease detection [24], [25]. Expert Systems and the Transition to Web-Based Platforms The knowledge-based systems of the 1980s–1990s, often referred to as expert systems, were the forerunners of today’s fuzzy-based healthcare tools. These systems encoded the expertise of clinicians into rule bases that could be consulted to assist in medical decision-making [26]. Scholar J Computational Science 93 Initially developed as standalone software, expert systems had several limitations, including restricted accessibility, high maintenance costs, and the inability to integrate with evolving medical guidelines. However, with the advancement of the internet, cloud computing, and graphical user interfaces (GUIs), modern fuzzy logic applications are increasingly implemented as web-based and mobile platforms. Such platforms provide broader accessibility, real-time updates, and scalability for large populations [18]. Fuzzy Logic in Breast Cancer Risk Prediction Breast cancer risk assessment exemplifies a domain where fuzzy logic’s strengths directly address clinical challenges. Risk factors such as age at menarche, body mass index (BMI), alcohol intake, duration of breastfeeding, and reproductive history often exhibit continuous or subjective characteristics that are poorly captured by threshold-based models [14]. Fuzzy logic offers the ability to:  Translate these linguistic descriptors into quantitative membership functions (e.g., “early menarche,” “high BMI”)  Combine multiple heterogeneous risk factors through IF–THEN rules (e.g., IF age at first pregnancy is late AND BMI is high THEN risk is high)  Provide human-interpretable reasoning that aligns with clinical decision-making [15], [16]. A notable trend is the development of multi-level fuzzy inference systems (FIS):  Yilmaz and Ayan [15] demonstrated that a Mamdani-based FIS could classify patients into five risk categories with 78.3% accuracy.  Latha [17] used fuzzy rules to analyze tumor-related data in a clinical decision-support environment.  Awodele et al. [16] introduced a mobile-based fuzzy expert system for preliminary breast cancer risk assessment, though it was limited to Android devices.  Gupta et al. [18] combined fuzzy rules with a decision-tree model to improve interpretability and diagnostic precision. Fuzzy Logic Models for Breast Cancer Risk Prediction Breast cancer risk assessment is inherently complex due to the heterogeneous interaction of genetic, hormonal, reproductive, and lifestyle factors. Unlike traditional regression models that use crisp thresholds, fuzzy logic systems excel at capturing continuous and linguistic variables, enabling nuanced stratification of risk levels. Over the last two decades, researchers have developed a variety of fuzzy inference models to support clinicians and empower patients with early risk-assessment tools. Scholar J Computational Science 94 Key Risk Factors Incorporated in Fuzzy Models Most fuzzy-based breast cancer risk prediction models integrate a combination of epidemiological and clinical input variables. These variables are typically normalized to the range [0, 1] using triangular, trapezoidal, or Gaussian membership functions (MFs). Commonly used factors include: Category Risk Factor Clinical Significance Demographic Age Incidence increases with age; women >50 years are at higher risk [27]. Reproductive Age at first menstrual cycle (AFMC), age at last menstrual cycle (ALMC), age at first pregnancy (AFP), age at menopause Early menarche (<12 yrs) or late menopause (>50 yrs) increase cumulative estrogen exposure and elevate risk [28]. Lactation history Duration of breastfeeding (DBF) Long-term breastfeeding (>12 months) has a protective effect [29]. Body composition Body mass index (BMI) Post-menopausal obesity is associated with increased risk [30]. Lifestyle Alcohol intake, smoking status, physical inactivity Alcohol is a dose-dependent risk factor; smoking adds moderate risk [31]. Family/Genetic Family history, BRCA mutation status High-penetrance genes contribute to hereditary breast cancer susceptibility [32]. Clinical/Pathologic Tumor histology, hormone-receptor status (in diagnostic/prognostic models) Incorporated mainly in diagnostic rather than early-risk models. These variables are defined as linguistic terms such as very young, young, average, old, very old or low, moderate, high. Each term is associated with a membership function that maps crisp input values to degrees of membership, enabling smooth transitions between categories. Scholar J Computational Science 95 Model Architectures Most breast-cancer fuzzy models use the Mamdani-type fuzzy inference system (FIS) because of its rulebased interpretability and ease of incorporating clinical expert knowledge [13]. A typical architecture includes: 1. Fuzzification layer: Converts crisp input values (e.g., age = 45) into degrees of membership for defined linguistic terms (middle-aged, young, etc.). 2. Rule base / Knowledge base: A set of IF–THEN rules derived from clinical literature and expert opinions. Example: 3. IF age is “old” AND BMI is “high” AND AFP is “late” THEN risk is “high.” 4. Inference engine: Aggregates rule outputs using fuzzy operators (min, max, product). 5. Defuzzification: Converts the fuzzy aggregated output into a crisp numerical risk score or a categorical label (e.g., low, moderate, high). Some models extend this architecture by integrating decision trees, neural networks, or hybrid optimization techniques to enhance performance or to learn membership functions automatically. Emerging Trends Recent fuzzy-based models show a shift from static desktop prototypes to weband cloud-based platforms, aiming to improve accessibility, scalability, and interoperability with electronic health records (EHRs). Some systems incorporate adaptive learning algorithms to automatically tune membership functions based on population data, addressing one of the long-standing criticisms of expert-elicited MFs. Furthermore, hybrid frameworks combining fuzzy inference with machine-learning algorithms (e.g., support-vector machines, neural networks, decision-trees) are increasingly popular for boosting predictive performance while retaining interpretability. Strengths and Remaining Challenges Strengths of fuzzy-based models:  Flexibility in modeling non-linear interactions of heterogeneous risk factors.  Ability to incorporate linguistic and expert-driven rules.  High interpretability compared to black-box models like deep learning.  Suitable for low-resource settings due to relatively low computational demand. Scholar J Computational Science 96 Challenges and gaps:  Subjectivity in MF design due to reliance on expert elicitation.  Limited external validation across diverse ethnic and demographic groups.  Small-scale datasets impede generalizability and clinical adoption.  Platform restrictions (e.g., Android-only) reduce universal accessibility.  Lack of integration with genomic and imaging biomarkers for comprehensive risk prediction. Future Research Directions Fuzzy-logic–based breast-cancer risk prediction has matured over the last two decades from prototype decision-support tools to mobile and web-based applications. However, most implementations remain proof-of-concept or locally validated rather than clinically deployed. To transform these promising models into robust, scalable, and clinically approved tools, several research and development priorities must be addressed. Development of Large-Scale, Diverse Benchmark Datasets A fundamental limitation of current fuzzy systems is the lack of standardized, population-wide datasets for model training, validation, and benchmarking. Future studies should:  Aggregate multi-center, multi-ethnic clinical cohorts with well-annotated epidemiological, reproductive, and lifestyle risk variables.  Establish open-access repositories that include both structured health records and longitudinal follow-up data to allow the assessment of predictive accuracy over time.  Include under-represented groups, such as women in lowand middle-income countries (LMICs), to improve equity and generalizability of risk models [2,4]. Hybrid Fuzzy–Machine-Learning Architectures While traditional Mamdani-type FIS models rely on expert-defined membership functions (MFs), modern approaches can use machine-learning (ML) to learn optimal MFs and generate new fuzzy rules automatically. Research priorities include:  Neuro-fuzzy systems that combine FL’s interpretability with the adaptive learning of neural networks [33]. Scholar J Computational Science 97  Evolutionary algorithms (e.g., genetic algorithms, particle-swarm optimization) for automatic MF tuning and rule-base pruning [34].  Hybrid fuzzy–decision-tree or fuzzy–support-vector-machine models to capture complex nonlinear patterns while maintaining transparent rule-based reasoning [18,35]. Integration of Multi-Modal and Personalized Data Most existing fuzzy models focus on epidemiologic and lifestyle risk factors. To improve predictive performance and clinical relevance, future systems should integrate:  Genetic and molecular biomarkers (e.g., BRCA1/2, TP53, polygenic risk scores) [32].  Digital-imaging phenotypes derived from mammography, ultrasound, MRI, and digital pathology, which capture structural and textural risk patterns [36].  Hormonal profiles, metabolic markers, and longitudinal weight-trajectory data, all of which influence lifetime estrogen exposure.  Patient-reported data on diet, exercise, reproductive history, and environmental exposures for comprehensive risk modeling. Such multi-modal integration can produce personalized, dynamic risk scores tailored to individual biology and lifestyle. Real-World Validation and Clinical Trials To achieve clinical translation, fuzzy models must undergo:  Prospective, multi-center clinical trials to evaluate predictive accuracy, usability, and impact on early detection rates.  Comparative effectiveness studies against established statistical models (e.g., Gail, Tyrer-Cuzick) to demonstrate added value.  Cost-effectiveness analyses to justify integration into population-screening programs, particularly in LMICs where resources are constrained.  Regulatory compliance with international standards for software-as-a-medical-device (SaMD), such as ISO 13485 and FDA/EMA AI-guidance documents. Cloud-Enabled, Interoperable Platforms A major barrier to adoption has been platform restriction (e.g., Android-only or offline desktop prototypes). Future work should emphasize:  Cloud-based, cross-platform architecture (web + mobile) to allow ubiquitous access by clinicians, patients, and screening programs. Scholar J Computational Science 98  Integration with electronic-health-record (EHR) systems to enable automated data ingestion and longitudinal risk tracking.  Scalable APIs and privacy-preserving federated learning to meet data-protection regulations (e.g., GDPR, HIPAA) while facilitating collaborative research. Explainability and User-Centric Design Maintaining trust and transparency is crucial in clinical AI. Fuzzy logic already offers interpretability via its rule-based structure, but further research should:  Develop visual dashboards that clearly display how input factors contribute to an individual’s risk score.  Explore explainable-AI (XAI) add-ons to hybrid fuzzy-ML systems to retain interpretability when using learned rules.  Conduct human-factors studies to ensure user-friendly interfaces for diverse end-users—clinicians, nurses, public-health workers, and patients. Ethical, Equity, and Policy Considerations As fuzzy-based tools move toward real-world screening, researchers must address:  Bias mitigation to avoid disadvantaging minorities or low-income populations through skewed training datasets.  Transparent reporting of uncertainty ranges to prevent overconfidence in borderline predictions.  Policies for equitable access, especially in low-resource regions where breast-cancer burden is increasing fastest [4]. Overall, the future of fuzzy-logic–based breast-cancer risk prediction lies in interdisciplinary collaboration among clinicians, computer scientists, epidemiologists, ethicists, and policymakers. By leveraging large datasets, hybrid learning algorithms, cloud-based delivery, and rigorous clinical validation, the next generation of fuzzy models can evolve from academic prototypes into globally deployable precisionscreening tools. Conclusion Fuzzy-logic–based models offer a transparent and adaptable alternative to conventional breast-cancer riskprediction tools by capturing uncertainty, integrating linguistic variables, and supporting clinicianinterpretable decisions. Evidence from studies such as [15-16], and [18] demonstrates the feasibility of these systems for early risk stratification. However, their clinical impact remains limited by small validation