Early Detection of Emphysema Progression Vladlena Gorbunova1, Sander S.A.M. Jacobs2, Pechin Lo1, Asger Dirksen3, Mads Nielsen1,4, Alireza Bab-Hadiashar5, and Marleen de Bruijne1,6 1 Department of Computer Science, University of Copenhagen, Denmark 2 Department of Biomedical Engineering, Eindhoven University of Technology, The Netherlands 3 Department of Respiratory Medicine, Gentofte University Hospital, Denmark 4 Nordic Bioscience A/S, Herlev, Denmark 5 Swinburne University of Technology, Hawthorn, Australia 6 Biomedical Imaging Group Rotterdam, Erasmus MC, Rotterdam, The Netherlands Abstract. Emphysema is one of the most widespread diseases in sub- jects with smoking history. The gold standard method for estimating the severity of emphysema is a lung function test, such as forced expi- ratory volume in first second (FEV1). However, several clinical studies showed that chest CT scans offer more sensitive estimates of emphysema progression. The standard CT densitometric score of emphysema is the relative area of voxels below a threshold (RA). The RA score is a global measurement and reflects the overall emphysema progression. In this work, we propose a framework for estimation of local em- physema progression from longitudinal chest CT scans. First, images are registered to a common system of coordinates and then local im- age dissimilarities are computed in corresponding anatomical locations. Finally, the obtained dissimilarity representation is converted into a sin- gle emphysema progression score. We applied the proposed algorithm on 27 patients with severe emphysema with CT scans acquired five time points, at baseline, after 3, after 12, after 21 and after 24 or 30 months. The results showed consistent emphysema progression with time and the overall progression score correlates significantly with the increase in RA score. 1 Introduction Emphysema is one of the most common chronic obstructive pulmonary diseases [1]. It is characterized by irreversible destruction of the lung parenchyma and usually caused by smoking [2]. In clinical practice, the severity of emphysema is commonly assessed using different lung function tests. Along with the lung function tests chest CT scans has been used for diagnosis of emphysema and detection of emphysema progres- sion. The standard CT density scores, such as relative area (RA) below certain threshold, e.g. -950 HU or -930 HU, and the n-th percentile density (nPD) of the lungs, were applied to estimate the emphysema progression [3,4]. CT densitom- etry scores have shown to be more sensitive measures of emphysema progression than lung function tests [4]. T. Jiang et al. (Eds.): MICCAI 2010, Part II, LNCS 6362, pp. 193���200, 2010. c Springer-Verlag Berlin Heidelberg 2010

194 V. Gorbunova et al. One of the major drawbacks of the standard CT density scores is their depen- dency on the inspiratory level [5,6]. Another important drawback is the lack of sensitivity, since the emphysema progression could only be measured once the intensity of lung tissue decreases below the standard threshold. Texture analysis may resolve this problems. This issue was investigated in a recent study, where a texture-based classification approach was proposed as alternative to the stan- dard emphysema scores [7]. The results showed that the texture-based approach outperforms the RA scores in differentiating diseased from healthy subjects. Several studies proposed how to estimate disease progression from longitudinal CT scans [5,6]. Authors proposed a method where CT scans are first registered to a common framework and then emphysema progression is estimated based on the average intensity decrease between the two successive scans. In this paper, we propose a more general way of assessing emphysema progres- sion between a pair of images. First, images are registered to a common system of coordinates. Second, local image histograms at a given location are obtained and dissimilarity measures between the histograms are computed. Thirdly, a measure of progression at the given location is derived from the dissimilarity measures. Finally, an overall disease progression score between the two images is computed. This method is applied to detect emphysema progression in a longitudinal study of patients with Alpha-1 antitrypsin deficiency [4]. 2 Method In this section we describe in details the work flow of the algorithm. The first subsection (2.1) briefly describes the image registration method that was applied to establish the spatial correspondence between images. The following subsec- tion (2.2) presents how local dissimilarities were constructed. The last subsection (2.3) describes how the local disease progression score on subject level was de- rived from the set of local dissimilarity measures. 2.1 Registration The image registration framework presented in [6] is used to register the images to a common system of coordinates. The framework starts with a preprocessing step, where the lung fields are extracted from the CT scans and the background value is set to 0 HU. First, an a���ne transform is applied to correct for global deformations. Then a series of multi-resolution B-Spline transforms with decreas- ing grid resolution is applied to the a���nely registered images. Each transform was optimized using the stochastic gradient descent method. Finally, the moving image is deformed based on the obtained deformation field. To minimize the intensity differences in the fixed and moving images caused by the difference in respiratory level, the intensities of the deformed image are adjusted with respect to the Jacobian determinant of the deformation field as proposed in [6]. The baseline image I1 was set as the fixed image, and the four follow up images I2..5 were set as the moving images in the registration framework.