Illustration of the obstacles in computerized lung segmentation using examples

Abstract

Purpose: Automated lung volume segmentation is often a preprocessing step in quantitative lung computed tomography (CT) image analysis. The objective of this study is to identify the obstacles in computerized lung volume segmentation and illustrate those explicitly using real examples. Awareness of these "difficult" cases may be helpful for the development of a robust and consistent lung segmentation algorithm.

Methods: We collected a large diverse dataset consisting of 2768 chest CT examinations acquired on 2292 subjects from various sources. These examinations cover a wide range of diseases, including lung cancer, chronic obstructive pulmonary disease, human immunodeficiency virus, pulmonary embolism, pneumonia, asthma, and interstitial lung disease (ILD). The CT acquisition protocols, including dose, scanners, and reconstruction kernels, vary significantly. After the application of a "neutral" thresholding-based approach to the collected CT examinations in a batch manner, the failed cases were subjectively identified and classified into different subgroups.

Results: Totally, 121 failed examinations are identified, corresponding to a failure ratio of 4.4%. These failed cases are summarized as 11 different subgroups, which is further classified into 3 broad categories: (1) failure caused by diseases, (2) failure caused by anatomy variability, and (3) failure caused by external factors. The failure percentages in these categories are 62.0%, 32.2%, and 5.8%, respectively.

Conclusions: The presence of specific lung diseases (e.g., pulmonary nodules, ILD, and pneumonia) is the primary issue in computerized lung segmentation. The segmentation failures caused by external factors and anatomy variety are relatively low but unavoidable in practice. It is desirable to develop robust schemes to handle these issues in a single pass when a large number of CT examinations need to be analyzed.

Figure 1

Illustration of the thresholding-based lung segmentation scheme used in this study: (a) a CT examination, (b) the result after the application of a thresholding operation, (c) the result after nonhuman body region removal, (d) the result after human body removal, (e) the result after trachea removal, (f) right and left lung labeling, (g) vessel inclusion, and (h) three-dimensional lung volume.

Figure 2

An example in the failure category “C1.” The indicated juxtapleural nodule is missed.

Figure 3

An example in the failure category “C2,” where the right and left lungs are attached when a thresholding procedure is applied.

Figure 4

An example in the failure category “C3.” In this example, the presence of mucus makes part of the left major bronchus not depicted on CT images. (a) and (b) Two neighboring image slices of the CT examinations. It can be seen that the left major bronchus is depicted on a slice (a) but disappear on the next neighboring slice (b).

Figure 5

An example in the failure category “C4.” In this example, the improper field-of-view makes the algorithm fail to identify the parenchyma regions, because the parenchyma is fused with the external nonlung regions in space.

Figure 6

An example in the failure category “C5.” In this example, the left lung was removed because of the existence of lung cancer. When labeling the right and the left lungs, the computer algorithm fails because the left lung cannot be found.

Figure 7

An example in the failure category “C6.” In this example, the right and left lung are incorrectly labeled. The patient position/orientation information is not available in the DICOM header of this specific case.

Figure 8

An example in the failure category “C7” where the metal spray effect in (a) leads to the inclusion of some nonlung region as shown in (b).

Figure 9

An example in the failure category “C8.” A large part of the right lung is missed due to the presence of severe pneumonia.

Figure 10

An example in the failure category “C9.” A large part of the right lung is missed due to the presence of severe interstitial lung disease (ILD).

Figure 11

An example in the failure category “C10.” In this example, there is an obvious Chilaiditi syndrome, or presence of gas in the right colic angle between the liver and the right hemidiaphragm. Due to its attachment to the right lung in image space, the gas region in hepatic flexure of colon is incorrectly included as part of the right lung.

Figure 12

An example in the failure category “C11,” where a slice of image is damaged. This issue causes the fusion of the lung region with the outside nonlung region, ultimately leading the failure in lung volume identification.