Smart microscopy: machine learning-guided adaptive label-free multimodal microscope with QPI and SLAM

We developed an adaptive microscopy platform that integrates quantitative phase imaging (QPI) with simultaneous label-free autofluorescence multi-harmonic (SLAM) microscopy, utilizing machine learning-guided decision-making. The system performs rapid, label-free QPI screening to extract morphological features based on phase reconstruction. These quantitative features feed into a trained neural network that identifies regions of biological interest and predicts dynamic cellular events in real-time. Upon detecting significant features or impending phenomena, such as mitosis or cancerous regions, the platform automatically switches to SLAM imaging for targeted, high-resolution acquisition of selected regions of interest. SLAM provides key molecular specificity and functional readouts, including harmonic generation and metabolic imaging capabilities. By intelligently allocating imaging modalities based on continuous computational analysis, our platform optimizes data acquisition efficiency while ensuring critical biological events are captured appropriately. This hierarchical imaging strategy represents a paradigm shift from static to dynamic microscopy workflows.