Fluctuation-Based Super-Resolution Microscopy Classification via Gradient Boosting Decision Trees

Fluorescence fluctuation-based super-resolution microscopy has broad applications in observing subcellular structures and monitoring their kinetic processes. Therefore, it is highly demanded to systematically study the reconstruction quality of multiple fluctuation-based super-resolution algorithms under different fluorescence temporal fluctuations. In this study, the performances in the image quality of multiple super-resolution techniques under different conditions are quantitatively analyzed and compared by employing comprehensive evaluation metrics. Furthermore, a gradient boosting decision tree (GBDT) model was constructed to accurately predict the most suitable super-resolution algorithm based on input features encompassing frame numbers, on-off brightness, on-state probability, and signal-to-noise ratios. The results show that high-quality super-resolution images can be obtained by increasing image frames together with enhancing the fluorescence fluctuation signals, and the GBDT model demonstrates robust predictive capability, achieving high classification accuracy after iterative training. This study could facilitate rapid selection and implementation of fluctuation-based super-resolution techniques for subcellular organelle research under diverse fluorescent labeling conditions.