Texture-Based Identification and Characterization of Interstitial Pneumonia Patterns in Lung Multidetector CT

Abstract

Identification and characterization of diffuse parenchyma lung disease (DPLD) patterns challenges computer-aided schemes in computed tomography (CT) lung analysis. In this study, an automated scheme for volumetric quantification of interstitial pneumonia (IP) patterns, a subset of DPLD, is presented, utilizing a multidetector CT (MDCT) dataset. Initially, lung-field segmentation is achieved by 3-D automated gray-level thresholding combined with an edge-highlighting wavelet preprocessing step, followed by a texture-based border refinement step. The vessel tree volume is identified and removed from lung field, resulting in lung parenchyma (LP) volume. Following, identification and characterization of IP patterns is formulated as a three-class pattern classification of LP into normal, ground glass, and reticular patterns, by means of k-nearest neighbor voxel classification, exploiting 3-D cooccurrence features. Performance of the proposed scheme in indentifying and characterizing ground glass and reticular patterns was evaluated by means of volume overlap (ground glass: 0.734 +/- 0.057, reticular: 0.815 +/- 0.037), true-positive fraction (ground glass: 0.638 +/- 0.055, reticular: 0.942 +/- 0.023) and false-positive fraction (ground glass: 0.361 +/- 0.027, reticular: 0.147 +/- 0.032) on five MDCT scans.

Three-dimensional solid texture analysis in biomedical imaging: Review and opportunities

“…Texture-based identification and characterization of interstitial pneumonia patterns and diffuse lung diseases in MDCT has been investigated by several research groups during the past 5 years (Xu et al, 2005;Xu et al, 2006b;Xu et al, 2006a;Fetita et al, 2007a;Fetita et al, 2007b;Korfiatis et al, 2008;Boehm et al, 2008;Chang-Chien et al, 2009;Mariolis et al, 2010a;Korfiatis et al, 2010;Mariolis et al, 2010b;van Rikxoort et al, 2011). Interstitial and diffuse lung diseases create large regions of altered tissue (e.g., fibrosis, ground glass, emphysema, micronodules, consolidation) with well-defined texture properties, which are best described in terms of texture properties (Webb et al, 2001).…”

A Bag-of-Features Approach to Classify Six Types of Pulmonary Textures on High-Resolution Computed Tomography

“…An interesting work is given in the work [12], where irregular-shaped 2D ROIs clustered on homogenous textures show that they are more powerful than the 2D square-shaped ROIs. For the classifiers, all of the works adopt mature techniques, such as the Bayesian classifier [4], [13], [14], the artificial neural networks (ANN) [5], [9], the nearest neighbor (NN) classifier [6], [11], [12], [15], [16], and the support vector machine (SVM) [6]- [8], [10]. The types of pulmonary textures are usually determined by the purpose of CAD systems.…”

“…For CAD systems on a specific purpose, the number of textural type is relatively few. For example, only two types of textures (mild and normal) are considered for the detection of early interstitial lung diseases [9], two kinds of emphysema textures and normal textures are considered for chronic obstructive pulmonary disease (COPD) [10], [11], and a subset of three kinds of DLD textures is considered for interstitial pneumonia [15].…”

“…If features are more discriminative, classification becomes easier. Features can be designed by using one or several of the following measures on gray values [9], [15], such as statistical measures on histogram distributions, measures on gray-level co-occurrence matrices (GLCM) and measures on gray-level run-length matrices (GLRLM). More powerful features can be designed if these measures are combined with some other cues on geometrical information, such as fractal features [4], [13], [14] or shapes [7], [8].…”

Using Multiscale Visual Words for Lung Texture Classification and Retrieval

“…New imaging techniques provide useful information to radiologists and other clinicians leading to an accurate diagnosis without a need for invasive techniques such as biopsies [20]. As a consequence, radiologists face changing problems: several studies have shown a high interand intra-observer variability in image-based diagnosis but a reduced variability for experienced observers [2,11]. Quick access to similar cases of the past with accurate diagnosis and further case information appears as useful to less experienced radiologists in order to make decisions consistent [1].…”