A Graph Neural Network Framework for Breast Cancer Risk Prediction
T. Padma, Vinodkumar Kakde, Aruna Kumar Kavuru, TamilSelvi Madeswaran, Suad Al Riyami, Sanjay Gupta · 2025
Breast cancer prediction is a critical task in medical diagnostics, and various machine learning techniques have been employed to improve the accuracy of such predictions. In this study, a methodology for solving the breast cancer prediction problem using Graph Neural Networks (GNNs) has been proposed. The conventional tabular format of datasets is transformed into a graph structure, where each data point is represented as a node. Edges are then created based on the similarities between the features of the data points using K-Nearest Neighbours (KNN). Each node's features represent diagnostic attributes such as radius, texture, and area, while the edges capture the relationship between similar samples. This graph-based representation allows the model to capture the underlying relationships and dependencies between the data points, which may not be apparent in a traditional feature-based approach. The proposed prediction model utilizes a Graph Convolutional Network (GCN) to learn the underlying patterns in the data and classify tumours as benign or malignant. GNNs are ideal for medical diagnosis as they effectively model complex, non-linear relationships between data points, which traditional methods may overlook. The model is trained on the WBC dataset, and its performance is evaluated based on classification accuracy, demonstrating the effectiveness of GNNs in capturing complex dependencies within medical data. The results highlight the potential of GNNs to enhance predictive models in the context of breast cancer diagnosis, offering a promising alternative to traditional machine learning techniques by leveraging graph-based data representations.