ALTIS

ALTIS segments lungs and trachea in computed tomography (CT) scans to provide robust anatomical delineation and quantitative analysis of pathological variations.

Key Features:

• Robustness Against Abnormalities: ALTIS uses image features and relative shape- and intensity-based characteristics that are less susceptible to appearance changes caused by disease, trauma, or prior surgical interventions.
• Efficiency: The method is optimized for speed, completing segmentation tasks within seconds on modern personal computers.
• High Accuracy: On a dataset of 1255 CT scans including numerous anomalous cases, ALTIS outperformed SOSM-S, MALF, and PTK, achieving average DICE coefficients of 0.987 for both lungs and 0.898 for the trachea and Average Symmetric Surface Distances (ASSD) of 0.938 (right lung), 0.856 (left lung), and 1.316 (trachea).

Scientific Applications:

• Abnormality detection and quantification: Detection, segmentation, and quantitative analysis of abnormal structures within lung parenchyma and the pleural space from CT scans.
• Clinical and research imaging studies: Analysis of anatomical variations across diverse patient populations and pathological conditions for clinical evaluation and research.

Methodology:

ALTIS applies a sequence of Image Foresting Transforms (IFTs) in three steps: lung-and-trachea extraction; internal seed estimation for background, trachea, left lung, and right lung; and delineation via optimum-path forests where each object is defined by an optimum-path forest rooted at its internal seeds.