RAI Labs Duke
Breast CAD | Chest CAD | Lung Dis CT | XCAT-MOBY-ROBY | Breast Models | FE Cardiac Model
Quant. Image | Emerg. Quant. Imaging | Perf. Metrology | Clinical Trials | Emerg. Clinical

Computerized quantification of interstitial lung diseases on CT

Quantification of diffuse interstitial lung disease on CT is an important and difficult clinical task. Radiologists often need to estimate the extent and progression of the interstitial lung disease in a patient because the extent and progression of the disease are reliable indicators for predicting treatment efficacy and patient survival. A computer-aided diagnostic system can help radiologists improve their accuracy and efficiency in quantifying the disease extent and progression. Because disease extent is defined as the ratio of disease areas to lung areas, the accurate segmentation of lungs and disease areas are two key techniques.

Thresholding of CT values was used to obtain initial segmentation results for lungs. In lung with severe interstitial disease, the initial segmentation results often failed to identify the parenchyma. Two texture features were employed to segment lungs with severe interstitial lung disease in CT. Figure 1 shows (a) the original CT slice, (b) initial lung segmentation by use of CT values only, and (c) final lung segmentation by using texture information.  It is apparent that texture information is very useful in segmenting lungs with severe abnormality.

  (a)   (b)   (c)  
Fig. 1. Segmentation of lungs with severe interstitial lung diseases.

Texture features extracted from run-length matrices and co-occurrence matrices were employed to further identify the disease areas inside the segmented lungs. Figure 2 shows (a) the original CT slice with interstitial lung disease, (b) initial disease segmentation by use of texture features, and (c) final disease segmentation by use of texture features and context information. The white dots indicate the regions of interest (ROIs) that are classified as “abnormal” by the computerized quantification scheme. It is apparent that the context information is very useful for removing false-positive ROIs.

  (a)   (b)   (c)  
Fig. 2. Segmentation of disease areas inside lungs.
Radiographic Tech. | Low-Cost Tomo | Optimiz. of Rad. Ther. | Personal CAD | Adapt. Educ.