TLDR: A new expert system leverages fuzzy soft set theory to assess breast cancer risk using clinical and physiological parameters like BMI, insulin, leptin, adiponectin, and age. The system fuzzifies data, transforms it into soft sets, reduces parameters, and applies Kong’s algorithm to predict risk. Developed and validated with data from the UCI Machine Learning Repository, the model achieved 70% accuracy in identifying high-risk patients, aiming to support healthcare professionals in early detection and the necessity of further diagnostic procedures.
Breast cancer remains a significant health challenge globally, being one of the leading causes of death among women. Early and accurate diagnosis is crucial for effective treatment and improving survival rates. However, the complex nature of the disease and the variability in patient risk factors make timely detection difficult. To address this, a new study introduces an expert system based on fuzzy soft set theory designed to assess breast cancer risk in patients. This system uses common clinical and physiological measurements to provide a preliminary risk assessment, aiming to support healthcare professionals in identifying individuals who might need further diagnostic procedures like biopsies.
The proposed expert system integrates several key input variables: Body Mass Index (BMI), Insulin Level (IL), Leptin Level (LL), Adiponectin Level (AL), and age. These parameters are readily obtainable from routine blood analyses, making the assessment method non-invasive and accessible. The system processes these inputs through a series of fuzzy inference rules and soft set computations to estimate breast cancer risk. The dataset used for developing and validating this model was sourced from the UCI Machine Learning Repository.
Understanding the Methodology
The development of this soft expert system involves several structured steps. First, the raw patient data undergoes a process called ‘fuzzification’. This is necessary because the initial numerical data is not directly suitable for soft sets. Fuzzification transforms these precise numerical values into ‘fuzzy sets’, which represent degrees of membership rather than absolute categories. For instance, age is categorized into fuzzy sets like ‘Child’, ‘Young’, ‘Mild’, and ‘Old’, each with a specific range and a membership function that determines how strongly a particular age belongs to that category.
Similarly, BMI is divided into ‘Obesity Class I’, ‘Obesity Class II’, and ‘Obesity Class III’. Insulin levels are classified as ‘Hypoglycemia’, ‘Normal’, and ‘Hyperinsulinemia’. Leptin is broken down into ‘Low-Leptin’, ‘Medium-Leptin’, ‘High-Leptin’, and ‘Very High-Leptin’, while Adiponectin is categorized into ‘Low-Adiponectin’, ‘Medium-Adiponectin’, and ‘High-Adiponectin’. Each of these fuzzy sets has defined membership functions, often trapezoidal or triangular, to quantify the degree to which a patient’s measurement falls into a particular fuzzy category.
After fuzzification, these fuzzy sets are transformed into ‘fuzzy soft sets’. This step combines the concepts of fuzzy set theory (dealing with vagueness) and soft set theory (dealing with uncertainty and parameterization). The system then performs ‘parameter reduction’, a crucial step that efficiently reduces the number of parameters by focusing only on those critical for decision-making, ensuring that the final decision is not influenced by less significant factors.
Finally, Kong’s algorithm is employed to analyze the processed data and forecast which patients are likely to develop breast cancer. The system generates a ‘degree risk factor’ for breast cancer for each patient.
Also Read:
- Uncovering Age Bias in Medical Image Segmentation: A Deep Dive into Label and Representational Disparities
- Benchmarking Privacy and Performance in Federated Tumor Segmentation
Results and Implications
The study evaluated the system’s performance using a sample of ten patients from the UCI Machine Learning Repository. The results are presented in a score table, where a higher score indicates a greater risk of breast cancer. The model correctly identified 7 out of 10 patients, achieving an accuracy of 70%. Specifically, patients μ3, μ11, μ19, μ31, and μ45 were classified as healthy controls, while patients μ60, μ71, μ82, μ91, and μ104 were identified as high-risk. The model correctly predicted the status of μ3, μ11, μ19, μ71, μ82, μ91, and μ104, but misclassified μ31, μ45, and μ60.
This fuzzy soft set theory-based expert system demonstrates the potential of soft computing techniques in expanding automatic breast cancer diagnostic regimes. It highlights the ability to use both linguistic and non-linguistic data, offering rapid and accurate learning capabilities. The authors emphasize that such soft expert systems, being rule-based, can quickly build upon existing connections and record-based measures to assist medical professionals in making timely decisions. This innovative research, detailed in the full paper available here, paves the way for future advancements in medical diagnosis.
Future research directions include extending this method using Fuzzy Rough Nearest Neighbors classification and prediction paradigms to better handle inherent uncertainty within datasets. This could involve combining the technique with neural network algorithms to classify or predict decision values by leveraging lower and upper approximations from fuzzy-rough set theory.
