Deep Learning-Driven Pattern Analysis of Dried E. coli-Laden Urine Deposits for Point-of-Care Diagnostics

Urinary tract infections (UTIs) are a rising global health concern, primarily caused by Escherichia coli (E. coli), disproportionately affecting women and the elderly. Despite advancements in diagnostic techniques, critical gaps persist: time delays, high costs, and false-positive predictions. This necessitates the development of rapid, reliable, and low-cost point-of-care diagnostic tools. As a vital step towards addressing this need, we present a machine learning-based morphological pattern analysis framework that evaluates dried deposits formed by E. coli-laden urine droplets. Following controlled evaporation, urine samples inoculated with E. coli at three distinct concentrations are imaged via brightfield microscopy. The computational objective of this study is twofold. First, we perform supervised ternary classification of the deposits based on bacterial concentration, utilizing a lightweight deep convolutional backbone strictly as a spatial feature extractor. Furthermore, to visually analyze feature cluster distributions, the standardized high-dimensional feature embeddings are projected using PCA and t-SNE. Second, beyond pattern classification, we define a Severity Factor by applying internal cluster validity indices (CVIs) to these extracted feature embeddings to evaluate morphological pattern deviation. Ultimately, this study establishes a proof-of-concept methodology for analyzing E. coli-laden dried deposits, offering a robust foundation with potential for integration into cyber-physical systems and real-time point-of-care diagnostics in resource-limited settings.