Tsebesebe, Nkgaphe TMpofu, Kelvin TSivarasu, SMthunzi-Kufa, P2026-01-052026-01-0520250-7988-5673-4http://hdl.handle.net/10204/14563In clinical investigation of biomolecular interactions, surface plasmon resonance imaging (SPRi), a labelfree detection technique, has proven to be an appropriate and dependable platform. A useful option for biosensing platforms is surface plasmon resonance-based sensors because of their excellent sensitivity and fine resolution. However, the detection limit of the SPRi-based detection method is limited by its relatively poor signal-to-noise ratio. Machine learning (ML) approaches therefore have the potential to address these problems. Machine learning techniques can help create SPR sensors more efficiently and increase their performance by automating the diagnostic process and increasing the signal-to-noise ratio. In this work, SPRi images are used to train the k-nearest neighbor algorithm for image classification. The KNN model correctly predicted all SPRi images in the dataset by achieving 100% accuracy, sensitivity, and specificity. With an AUC of 1.0, the model performed excellently, suggesting exceptional potential for automated real-time detection of binding events. Therefore, the machine learning-enhanced SPRi platform has the potential to offer a robust alternative to conventional manual interpretation in SPR sensing.FulltextenSurface plasmon resonance imagingMachine learningClassification bindingNonbindingArticleN/A