ilastik

ilastik performs machine-learning-based bioimage segmentation, classification, counting, and tracking for multi-dimensional microscopy data.

Key Features:

• Workflow adaptation via sparse annotations: Supports customization of pre-configured workflows through sparse training annotations to tailor classifiers to specific imaging tasks.
• Nonlinear classification: Employs nonlinear classifiers trained on user-provided annotations to improve segmentation and classification accuracy in complex biological datasets.
• Multi-dimensional data processing: Processes up to five dimensions (3D spatial, time, and multiple channels) for dynamic and multi-channel imaging data.
• On-demand computational backend: Uses an on-demand computational backend that performs operations as needed, enabling predictions and processing of datasets larger than available RAM.
• Command-line execution of workflows: Allows trained workflows to be applied to new datasets via command-line execution to support high-throughput processing.

Scientific Applications:

• Segmentation and classification of biological structures: Generates pixel- or object-level segmentations and classifications for diverse biological imaging tasks.
• Object counting and tracking: Supports counting and tracking of cells or subcellular structures in time-lapse microscopy datasets.
• Multi-channel fluorescence and advanced imaging: Handles multi-channel fluorescence imaging and other advanced imaging modalities requiring multi-dimensional analysis.

Methodology:

Training of machine-learning models using user-provided sparse annotations and nonlinear classifiers; on-demand computational backend performs operations as needed to enable processing of datasets larger than available RAM; trained workflows can be executed via command line on new datasets.