LibMI

LibMI provides efficient reading, writing, modification, and annotation of whole-slide images (WSIs) using the Modifiable Multiscale Image (MMSI) format to support large-scale histopathological image processing.

Key Features:

• Open and standardized image file format: Implements the Modifiable Multiscale Image (MMSI) format enabling reading, writing, modification, and annotation of whole-slide images (WSIs).
• Efficient data management: Uses a pyramid image structure with lazy propagation from segment tree algorithms to achieve linear time complexity for read and modify operations.
• Caching mechanism: Incorporates a cache to accelerate repeated access and processing of image regions.
• Pixel-wise data storage: Supports storing and retrieving pixel-wise information on WSIs for detailed annotation and analysis.
• Implementation and bindings: Core library implemented in C++ with Python bindings for integration into analysis workflows.
• Performance benchmark: Demonstrated 18.237 times faster read-only performance compared with similar libraries and file formats.

Scientific Applications:

• Image thresholding: Facilitates segmentation and analysis of specific regions within WSIs.
• Microscopic color correction: Provides methods to correct color variation in microscopic images to improve visual and quantitative analysis.
• Pixel-wise annotation and storage: Enables storage and retrieval of pixel-level information on WSIs for detailed annotation and downstream analysis.

Methodology:

Implements the MMSI format with a pyramid image structure, lazy propagation via segment tree algorithms for linear-time read/modify operations, and a caching mechanism; core implemented in C++ with Python bindings.