MIST

MIST stitches large-scale microscopy image tiles to assemble accurate 2D time-lapse mosaics for specimens that exceed a single microscope field of view.

Key Features:

• Automated Stitching: Calculates pairwise translations between overlapping image tiles to assemble seamless mosaics.
• Mechanical Stage Model Optimization: Estimates mechanical stage parameters including actuator backlash and stage repeatability ('r') to refine alignment.
• Translation Optimization Area: Optimizes translations within a square area of (4r)2 to minimize errors from mechanical inconsistencies.
• High-Performance Computing: Implements multicore hybrid CPU/GPU computation for high-throughput processing of large datasets.
• Throughput Performance: Processes terabytes of time-lapse multi-channel mosaics 15–100× faster than existing tools.
• Reference Datasets: Validated on 15 reference datasets including stem cell colonies with overlap percentages from 10% to 50%.
• Accuracy Metrics: Achieves an average centroid distance error of less than 2% of a field of view and outperforms three other open-source tools in precision.

Scientific Applications:

• Time-Lapse Studies: Facilitates observation and quantitative analysis of entire cell cultures over time across multiple imaging modalities.
• Large-Scale Specimen Imaging: Enables assembly and analysis of specimens that exceed the field of view, such as tissue sections and large cellular aggregates.

Methodology:

Computational steps include calculating pairwise translations between overlapping tiles, estimating mechanical stage model parameters (including actuator backlash and stage repeatability 'r'), and optimizing translations within the (4r)2 area to reduce stitching errors from mechanical uncertainties, photobleaching, segmentation issues, and inherent stitching inaccuracies.