Radiological evaluation of interstitial lung disease

Chest radiograph, computed tomography and high-resolution computed tomography (HRCT) are an integral part in the diagnosis and evaluation of diffuse interstitial lung disease (DILD) and are reviewed briefly. This review then delves into the burgeoning interest associated with enhanced imaging and computational capabilities of multidetector HRCT in the automated evaluation of computer-aided recognition of different patterns associated with DILD, automated regional and global distribution and quantitation of DILD, and automated longitudinal comparison of HRCT scans to estimate progression of DILD.

Problems associated with automatic segmentation of lung in DILD because of high-attenuation lesions in subpleural locations have been overcome to a large extent. Use of two-dimensional HRCT has made it possible to automate identification and quantification of patterns associated with DILD with an overall accuracy of 89%. Significant progress has been made with automated image registration and noise reduction, which are essential for sequential comparison of HRCT in DILD. Early experience with automated determination of progression or regression of DILD on two-dimensional HRCT, using the dissimilarity-based feature strategy, compared favorably with radiologist assessment and approached a concordance rate of 80%.

The combination of multidetector HRCT, the support vector machine classifier and various computational algorithms have made it possible to fully automate isolation of lungs in DILD, recognize radiographic patterns associated with DILD and sequentially estimate progression of DILD. It is hoped there will be further improvement when these are extended to three-dimensional HRCT.