Nonparametric Variational Infinite Libby-Novick Beta Mixture Model for Medical Data Clustering

Unsupervised clustering techniques are essential for extracting meaningful structure from complex imaging datasets across biomedical and scientific domains. We propose a Variational Infinite Libby-Novick Beta Mixture Model (VILNBMM), a framework that automatically determines the optimal number of clusters in imaging data by combining Dirichlet process priors with the Libby-Novick Beta (LNB) distribution. Our approach handles complex distributional patterns in bounded domains without requiring the presetting of component numbers a priori. We develop a variational inference algorithm that reformulates posterior estimation as an optimization problem, making the model computationally tractable for real world ap-plications. Experimental evaluation on medical imaging datasets demonstrates the effectiveness of VILNBMM for unsupervised Intelligent diagnostic support systems.