An Analysis of Paramecium bursaria ToF‐SIMS/OrbiSIMS Fusion Data via Unsupervised Machine Learning

Identification and imaging of target materials in complex biological samples are crucial for evaluating cells and tissues. Time‐of‐flight secondary ion mass spectrometry (ToF‐SIMS) provides detailed chemical information and material images with a spatial resolution of approximately 100 nm. However, its mass‐resolving power and mass accuracy are sometimes insufficient to identify peaks even below a few hundred Da. In contrast, OrbiSIMS has at least an order of magnitude higher mass‐resolving power and mass accuracy than ToF‐SIMS, enabling more confident annotation of mass peaks. However, its spatial resolution is lower than that of ToF‐SIMS using a liquid metal ion gun (LMIG) because OrbiSIMS generally employs a gas cluster ion beam (GCIB) for organic imaging. Image data fusion utilizes the advantages of both techniques: the high spatial resolution of ToF‐SIMS and the high mass resolution of OrbiSIMS. Applying unsupervised machine learning to fused ToF‐SIMS and OrbiSIMS data associates mass peaks from both techniques with the same material. ToF‐SIMS and OrbiSIMS datasets of Paramecium bursaria with symbiotic algae, Chlorella variabilis , were analyzed to evaluate this approach. Mass peaks from biological tissues were identified using OrbiSIMS masses, with detailed distributions provided by ToF‐SIMS secondary‐ion images. This study demonstrates that combining image fusion and machine learning enables chemical localization of organelles, including the nucleus and Chlorella cells, with both high mass and spatial resolution.