Discerning the cellular response using statistical discrimination of fluorescence images of membrane receptors

Abstract

We demonstrate a statistical modeling technique to recognize T cell responses to different external environmental conditions using membrane distributions of T cell receptors. We transformed fluorescence images of T cell receptors from each T cell into estimated model parameters of a partial differential equation. The model parameters enabled the construction of an accurate classification model using linear discrimination techniques. We further demonstrated that the technique successfully differentiated immobilized T cells on non-activating and activating surfaces. Compared to machine learning techniques, our statistical technique relies upon robust image-derived statistics and achieves effective classification with a limited sample size and a minimal computational footprint. The technique provides an effective strategy to quantitatively characterize the global distribution of membrane receptors under various physiological and pathological conditions.