Application of texture analysis to ventilation SPECT/CT data

Abstract

It is demonstrated that textural parameters calculated from functional pulmonary CT data have the potential to provide a robust and objective quantitative characterisation of inhomogeneity in lung function and classification of lung diseases in routine clinical applications. Clear recommendations are made for optimum data preparation and textural parameter selection. A new set of platform-independent software tools are presented that are implemented as plug-ins for ImageJ. The tools allow segmentation and subsequent histogram-based and grey-level co-occurrence matrix based analysis of the regions of interest. The work-flow is optimised for use in a clinical environment for the analysis of transverse Computed Tomography (CT) scans and lung ventilation scans based on SPECT. Consistency tests are made against other texture analysis plug-ins and simulated lung CT data. The same methods are then applied to patient data consisting of a healthy reference group and one patient group each who suffered from asthma, chronic obstructive pulmonary disease (COPD), and COPD plus lung cancer. The potential for disease classification based on computer analysis is evaluated.