Abstract: Novel systems and methods for airway segmentation are disclosed using confident airway volume, multi-scale topological and geometric leakage detection, exact leakage localization and correction, forbidden volume insertion and freezing, iterative shift in growing techniques starting with a conservative parameter/strategy and progressing toward generous ones, and a final pruning using machine learning, neural network, deep learning or artificial intelligence methods. The systems and methods are fully automated requiring no manual inputs or post-editing steps. The systems and methods use region growing-based segmentation to iteratively grow an airway tree starting with an initial seed, possibly inside the trachea, and a conservative segmentation parameter, e.g., a conservative threshold value for intensity-based region growing. The method then gradually progresses toward more generous values of the segmentation parameters until there is convergence.

Abstract: Novel systems and methods for airway segmentation are disclosed using confident airway volume, multi-scale topological and geometric leakage detection, exact leakage localization and correction, forbidden volume insertion and freezing, iterative shift in growing techniques starting with a conservative parameter/strategy and progressing toward generous ones, and a final pruning using machine learning, neural network, deep learning or artificial intelligence methods. The systems and methods are fully automated requiring no manual inputs or post-editing steps. The systems and methods use region growing-based segmentation to iteratively grow an airway tree starting with an initial seed, possibly inside the trachea, and a conservative segmentation parameter, e.g., a conservative threshold value for intensity-based region growing. The method then gradually progresses toward more generous values of the segmentation parameters until there is convergence.