A Review of Image Segmentation Methods for Lung Nodule Detection Based on Computed Tomography Images

Abstract

The detection and segmentation of lung nodules based on computer tomography images (CT) is a basic and significant step to achieve the robotic needle biopsy. In this paper, we reviewed some typical segmentation algorithms, including thresholding, active contour, differential operator, region growing and watershed. To analyse their performance on lung nodule detection, we applied them to four CT images of different kinds of lung nodules. The results show that thresholding, active contour and differential operator do well in the segmentation of solitary nodules, while region growing has an advantage over the others on segmenting nodules adhere to vessels. For segmentation of semi-transparent nodules, differential operator is an especially suitable choice. Watershed can segment nodules adhere to vessels and semi-transparent nodules well, but it has low sensitivity in solitary nodules.

Automated delineation of non‐small cell lung cancer: A step toward quantitative reasoning in medical decision science

“…Segmentation of these lesions was found to be less complicated compared to that of mediastinal masses, because it is only attached to a smooth surface like the chest wall, in comparison to complicated structures such as the esophagus, trachea, thymus, or aorta for a mediastinal lesion. Conventional boundary‐ and energy‐based delineation such as Canny edge detector and active contour, have failed to accurately segment pleural wall masses with very high false positive rates . This study has reported that the chest wall was also falsely segmented along with the nodule using these models.…”

“…Wang et al demonstrated that energy‐based models (level set, graph‐cut) and machine learning‐based models (U‐Net, CF‐CNN) have efficiently segmented solitary masses, with sensitivity ranging from 64%‐92% . Zheng et al investigated four classes of segmentation algorithms, namely, intensity‐based (Otsu), region‐based (region growing, watershed transform), edge‐based (Canny detector) and energy‐based (active contour); and have qualitatively shown that all methods yield between good to excellent results; except for the watershed transform, which had the highest false negative rate. We present two examples from our datasets, as depicted in Figures and , for the solitary mass segmentation from subjects 024 and 051.…”

“…Conventional boundary-and energy-based delineation such as Canny edge detector and active contour, have failed to accurately segment pleural wall masses with very high false positive rates. 49 This study has reported that the chest wall was also falsely segmented along with the nodule using these models. However, Chung et al improved this limitation by introducing the idea of circle and ellipse detection using Hough transform in combination with an active contour model.…”

“…They are also commonly referred to as vascularized masses. Region-based algorithms such as region growing and watershed transform exhibited excellent results in separating a vascularized mass from the bronchial branch, 49 since attenuation values between nodule tissue and its surrounding airway structures are distinct. A model based on boundary detection however is very likely to return high false positives, in which the airway branches tend to be included in the results.…”

A Coarse-to-fine Morphological Approach With Knowledge-based Rules and Self-adapting Correction for Lung Nodules Segmentation

“…The morphological methods including thresholding method [4,5], region growing [6,7], active contours [8,9], and graph-cut methods [10,11], are often based on pixel density; whereas deep learning-based methods [12,13] use neural networks to extract semantic information [14]. Although morphological techniques do not necessitate as much processing as DNN, their absence of deep semantic traits and 3D information might result in incorrect segmentation, particularly for juxtapleural [9], ground-glass, and small-sized nodules [15]. To put it another way, their effects are size [5] or type dependent.…”