LiveLattice: Real-time visualisation of tilted light-sheet microscopy data using a memory-efficient transformation...

Area of Science:

• Biomedical Imaging
• Computational Biology
• Microscopy Techniques

Background:

• Light-sheet fluorescence microscopy (LSFM) enables high temporal resolution 4D imaging.
• Tilted-sample-scan LSFM techniques like iSPIM and LLSM generate data requiring deskewing and rotation.
• Current preprocessing methods for LSFM data are computationally intensive and memory-demanding, limiting real-time analysis.

Purpose of the Study:

• To develop a memory-efficient algorithm for fast preprocessing of large LSFM datasets.
• To reduce data preprocessing time and enable real-time visualization of 4D microscopy data.
• To overcome computational challenges associated with tilted-sample-scan LSFM data.

Main Methods:

• Developed WH-Transform, a novel, memory-efficient transformation algorithm.
• Implemented GPU acceleration for rapid data processing.
• Benchmarked WH-Transform against conventional methods and existing software.

Main Results:

• WH-Transform reduces preprocessing time for large image stacks by over 10-fold.
• The algorithm demonstrates linear runtime complexity, outperforming cubic and quadratic methods.
• A 2 GB 3D volume was preprocessed in 3 seconds on a standard workstation with GPU.
• Real-time preprocessing and visualization were achieved for 4D datasets of human hepatocytes, lung, and brain organoids.

Conclusions:

• WH-Transform significantly accelerates LSFM data preprocessing, making it memory-efficient and computationally feasible.
• The algorithm enables real-time data visualization and analysis on standard hardware.
• This advancement revolutionizes biological imaging applications using LSFM and similar microscopes by improving usability and workflow.