Computerized assessment of pulmonary fissure integrity using high resolution CT

Abstract

Purpose: Knowledge of pulmonary interlobar fissure integrity is of interest in a number of clinical and investigational applications. The authors developed and tested a high resolution CT based automated computerized scheme for this purpose.

Methods: The fissure integrity assessment scheme consists of the following steps: (1) Fissure detection, (2) individual fissure identification, (3) fissure type determination, (4) "complete" interlobe surface estimation, and (5) fissure integrity estimation. For evaluation purposes, 50 anonymized chest CT examinations were ascertained and the complete and "incomplete" regions of the fissures of interest were manually marked by two experienced radiologists. After applying the scheme to the same examinations, differences among fissure percent completeness estimates based on the radiologists' manual markings and the automated computerized scheme were computed and compared.

Results: Average differences in estimated fissure percent completeness (integrity) between the results of the computerized scheme and that based on each of the two radiologists' markings were 6.88% +/- 5.86%, 9.57% +/- 7.77%, and 4.19% +/- 5.64% for the right major fissures, the right minor fissures, and the left major fissures, respectively. The differences between results based on radiologists' markings for the same fissures were 4.27% +/- 3.32%, 7.02% +/- 5.54%, and 4.23% +/- 4.93%, respectively. The difference among the three matched measurement sets for each fissure were statistically significant (Friedman's test, p < or = 0.005) but paired comparisons showed that much of the observed difference was related to inter-reader differences rather than reader-computerized scheme differences. Computerized estimates were correlated with each of the radiologist's estimates (Spearman, p < 0.0001).

Conclusions: While variability between readers-based estimates of fissure integrity was smaller than differences between the computerized scheme and each of the readers, the result reported here are quite encouraging in that the magnitude of these differences were in the same magnitude, demonstrating the feasibility of using a computerized scheme for this purpose.

Figure 1

Schematic diagram of the fissure integrity assessment approach. The shaded steps indicate the work developed and presented as a part of this investigation.

Figure 2

An example of our pulmonary fissure detection result. (a)–(c) are the original CT images viewed along the axial, coronal, and sagittal directions, (e)–(g) show the automated fissure detection results as an overlay, (d) shows the three-dimensional fissure surfaces, and (h) shows the smoothed version of (d) after applying the Laplacian smoothing algorithm. The arrows labeled as “1” indicate the incomplete fissure not depicted on the CT images and the arrows labeled as “2” indicate detected accessory fissures.

Figure 3

The estimated fissure surfaces after linear morphological filtering. (a)–(c) show the results after applying the developed morphological filter to the detected fissures in Figs. 2e, 2f, 2g, respectively, and (d) shows the smoothed three-dimensional fissures modeled using the MCA (Ref. 26).

Figure 4

Illustration of the progressive fissure decomposition scheme. (a) The originally detected fissures; (b) a set of separated surface patches after the removal of regions of pulmonary fissures with high curvature; (c) the final marked fissure regions with their unique identifications after applying a progressive decomposition strategy (Fig. 5); and (d) individually segmented fissures after a merging operation.

Figure 5

A flowchart of the fissure decomposition algorithm.

Figure 6

An example of individual fissure identification when applying the fissure classification procedure to the CT examination in Fig. 2. (a)–(c) show the identification of different types of fissures in image space and (d) shows a three-dimensional representation of the identified fissures in geometric space.

Figure 7

Visualization of the curvature of different fissure surfaces in Fig. 3d. The visualized curvature in (a)–(c) represent the larger of |k₁| and |k₂|. The arrows in (a) and (b) point to the regions connecting the major and minor fissures and the arrow in (c) point to the region connecting the accessory and minor fissures. The result of removing regions with high curvature is shown in (d).

Figure 8

Possible impact of the sampling strategy on implicit surface fitting. (a)–(c) show the fitted curves/surfaces for given sets of constraints points with the same space positions but different normal vectors. The extrapolated regions of these points are located outside the end points (i.e., points 1 and 8); (d) illustrates a uniform sampling strategy where surface regions are divided into a set of uniform grids and the constraints points are selected as the averaged (including positions and normal vectors) of the vertices with small curvatures within each grid.

Figure 9

Complete fissure estimation based on the uniform sampling strategy. (a)–(c) show the interlobe boundaries in image space and (d)–(f) show the interlobe boundaries in geometric space viewed from different perspectives.

Figure 10

An illustration of the projection based fissure integrity quantification. The irregular region represents a set of actually detected fissure regions, the smooth curve (or surface) in (a) represents the estimated complete fissure (or interlobe boundary), the arrows denote the positive/negative of the normal vectors at points on the complete fissure, the smooth curve (or surface) in (b) represents the overlapping regions after projecting neighboring detected fissure region onto the complete fissure surface.

Figure 11

Two examinations in which the estimation errors are caused by the undersegmentation of the pulmonary fissures due to presence of diseases (e.g., interstitial disease and severe emphysema). The top row shows the left lung of case no. 23 and the bottom row shows the right lung of case no. 30. The left column shows the sagittal views of the CT examinations. The middle column shows the detected fissures in overlay. The lobe segmentation results are shown in the right column.

Figure 12

An examination (case no. 45) in which the estimation error is caused by incorrect individual fissure identification due to a complex topology of the fissures in the right lung. (a) The sagittal view of the CT examination, (b) the detected fissures in overlay, (c) the lobe segmentation results, and (d) the three-dimensional model of the pulmonary fissures in the right lung are shown, respectively.