Fast and robust feature-based stitching algorithm for microscopic images.

Area of Science:

• Microscopy
• Computational Biology
• Image Processing

Background:

• High-resolution microscopy has a limited field of view, hindering comprehensive analysis of biological samples.
• Whole-slide imaging (WSI) captures multiple tiles for stitching, but faces challenges with repetitive tissue textures, non-informative backgrounds, and computational demands.

Purpose of the Study:

• To develop a fast and robust algorithm for stitching microscopic images (tiles) obtained from WSI.
• To address the limitations of existing stitching methods, particularly concerning speed and accuracy with complex biological samples.

Main Methods:

• Proposed the Fast and Robust Microscopic Image Stitching (FRMIS) algorithm utilizing pairwise and global alignment.
• Employed Speeded Up Robust Features (SURF) for feature extraction and matching, initially focusing on small overlapping regions for efficiency.
• Implemented a fallback to full overlapping region feature extraction when initial matching fails, and utilized weighted graph construction for global alignment.

Main Results:

• FRMIS demonstrated significantly faster stitching times compared to the Microscopy Image Stitching Tool (MIST) across various modalities.
• Achieved performance improvements of 481% (bright-field), 259% (phase-contrast), and 282% (fluorescence) over MIST.
• The algorithm proved robust to uneven illumination and repetitive tissue textures, reducing misalignment issues.

Conclusions:

• FRMIS offers a computationally efficient and accurate solution for stitching large numbers of microscopic image tiles.
• The algorithm enhances the study of biological samples by enabling seamless reconstruction of WSI data.
• FRMIS represents a significant advancement in microscopic image analysis, particularly for large-scale WSI datasets.