inviCRO, together with the University of New Mexico and the Lovelace Respiratory Research Institute have developed an imaging assay to assess the deposition of radiolabeled particles of varying properties in the respiratory tract of mice and rats. The image analysis team at inviCRO has developed an approach based on topographic thinning or "onion peeling" to quantify the distribution of the radiolabeled particles within the lungs.

The reliable, fully-automated 3D CT lung segmentation scheme consists of the following six steps: First, the small-animal body is segmented using a heuristic threshold based on pixel intensity. In the second step, bone is segmented using threshold segmentation inside the body. The third step includes computation of intensity threshold and region homogeneity measurements to segment the air. Then the features are used, including size and location, to identify lung tissue. The fourth step is to separate fused left and right lung using 3D region splitting and growing. The stop criterion increases the power of a region-grow algorithm by utilizing the “history” of growing. At the fifth step, the segmented left and right lung regions are refined to prune the trachea and remove the mediastinium using features, including size, location, shape, and compactness, etc.  Finally, morphological operations are used to fill the small holes in the left and right lung and smooth the irregular boundary. To adapt to the variability of the large data sets, robust customized features are used in classification.