CT grading of lung disease in lymphangioleiomyomatosis

Abstract

Objective: The aim of this study was to determine whether a CT-based method shows lung lesions, grades disease severity, and evaluates lung tissue in areas adjacent to or remote from cysts in patients with lymphangioleiomyomatosis (LAM), a cystic lung disease that may cause respiratory failure and death.

Materials and methods: Three hundred twenty-six CT examinations of 52 patients with LAM were studied. After the lungs had been divided into segments and images had been subdivided into texture blocks, a multidimensional feature vector was used to differentiate and group each texture block. Cysts were outlined, and texture around and away from cysts was analyzed. Sequential CT scans and pulmonary function test results were evaluated to assess the trend of change. Histopathologic examinations were performed of biopsy specimens from 45 patients.

Results: Differences in texture features between areas adjacent to and areas remote from the cysts were observed. The cyst score and sum entropy in areas around the cysts correlated with lung function (p<0.0001). Emphysematouslike changes in noncystic areas were identified in lung tissue of 31 of 45 patients.

Conclusion: A computational method that uses texture analysis and feature correlation can identify and quantify cystic areas where LAM exists and can detect abnormalities in areas near cysts. Pathologic data also show lung damage in areas adjacent to cysts. Several texture features correlate with lung function. Declines in lung function paralleled changes in texture features. In LAM, cystic changes alone may not define the extent of lung destruction.

Fig. 1—

Diagram shows methods used to evaluate texture of CT images. Given lung CT dataset, lung is first segmented and cystic lesions are detected before texture analysis of regions in vicinity of cysts and in normal-appearing lung areas at distance from them. Results of texture analysis are then compared with emphysematouslike changes.

Fig. 2—

Results of lung segmentation in 41-year-old woman with lymphangioleiomyomatosis. Input image (left) and segmented image (right) are shown. Green represents segmented left lung, and red represents segmented right lung.

Fig. 3—

50-year-old woman with lymphangioleiomyomatosis. Lymphangioleiomyomatosis cystic lesions are detected using histogram analysis and mathematic morphology. Cystic lesions identified by this procedure are shown in green.

Fig. 4—

Images were selected from multiple patients and show texture examples in normal-appearing lung regions at distance from cysts (top row) and in lung regions near cysts (bottom row). The two types of regions manifest distinct texture characteristics.

Fig. 5—

29-year-old woman with lymphangioleiomyomatosis (LAM). Photomicrograph of histologic section at scanning power microscopy shows two large LAM cysts, one of which has a small bundle of LAM cells in its wall (arrow). The surrounding lung shows dilated airspaces similar to those encountered in emphysema (asterisks).

Fig. 6—

Longitudinal analyses. A–D, Percent-predicted forced expiratory volume in 1 second (FEV₁) (A), diffusing capacity of lung for carbon monoxide (Dlco) (B), cyst scores (C), and sum entropy near cysts (D) in 11 patients with lymphangioleiomyomatosis. Time interval between each CT scan and physiologic data is approximately 1 year.