Body region localization in whole‐body low‐dose CT images of PET/CT scans using virtual landmarks

Abstract

Purpose Radiological imaging and image interpretation for clinical decision making are mostly specific to each body region such as head and neck, thorax, abdomen, pelvis, and extremities. In this study, we present a new solution to trim automatically the given axial image stack into image volumes satisfying the given body region definition. Methods The proposed approach consists of the following steps. First, a set of reference objects is selected and roughly segmented. Virtual landmarks (VLs) for the objects are then identified by using principal component analysis and recursive subdivision of the object via the principal axes system. The VLs can be defined based on just the binary objects or objects with gray values also considered. The VLs may lie anywhere with respect to the object, inside or outside, and rarely on the object surface, and are tethered to the object. Second, a classic neural network regressor is configured to learn the geometric mapping relationship between the VLs and the boundary locations of each body region. The trained network is then used to predict the locations of the body region boundaries. In this study, we focus on three body regions — thorax, abdomen, and pelvis, and predict their superior and inferior axial locations denoted by TS(I), TI(I), AS(I), AI(I), PS(I), and PI(I), respectively, for any given volume image I. Two kinds of reference objects — the skeleton and the lungs and airways, are employed to test the localization performance of the proposed approach. Results Our method is tested by using low‐dose unenhanced computed tomography (CT) images of 180 near whole‐body 18F‐fluorodeoxyglucose‐positron emission tomography/computed tomography (FDG‐PET/CT) scans (including 34 whole‐body scans) which are randomly divided into training and testing sets with a ratio of 85%:15%. The procedure is repeated six times and three times for the case of lungs and skeleton, respectively, with different divisions of the entire data set at this proportion. For the case of using skeleton as a reference object, the overall mean localization error for the six locations expressed as number of slices (nS) and distance (dS) in mm, is found to be nS: 3.4, 4.7, 4.1, 5.2, 5.2, and 3.9; dS: 13.4, 18.9, 16.5, 20.8, 20.8, and 15.5 mm for binary objects; nS: 4.1, 5.7, 4.3, 5.9, 5.9, and 4.0; dS: 16.2, 22.7, 17.2, 23.7, 23.7, and 16.1 mm for gray objects, respectively. For the case of using lungs and airways as a reference object, the corresponding results are, nS: 4.0, 5.3, 4.1, 6.9, 6.9, and 7.4; dS: 15.0, 19.7, 15.3, 26.2, 26.2, and 27.9 mm for binary objects; nS: 3.9, 5.4, 3.6, 7.2, 7.2, and 7.6; dS: 14.6, 20.1, 13.7, 27.3, 27.3, and 28.6 mm for gray objects, respectively. Conclusions Precise body region identification automatically in whole‐body or body region tomographic images is vital for numerous medical image analysis and analytics applications. Despite its importance, this issue has received very little attention in the literature. We present a solution to this problem in this study using the conce...

BRR‐Net: A tandem architectural CNN–RNN for automatic body region localization in CT images

“…Before examining the results presented henceforth, it is important to examine the variations in human performance in the task of labeling the body region boundaries. In a previous work 1 (Bai et al, 2019), the TS ( I ), TI ( I ), AS ( I ), AI ( I ), and PI ( I ) classes were labeled for 180 volumetric images by an expert. For this study, the images were labeled again, by a different expert.…”

“…As mentioned in Section 1.B, the only other works that tackled the problem addressed in this paper in a much‐limited sense are Bai et al 1 and Hussein et al 4 Bai et al focused on localizing TS ( I ), TI ( I ), AS ( I ), AI ( I ) = PS ( I ), and PI ( I ) in low‐dose CT of PET/CT acquisitions as well. Using a different concept of virtual landmarks and neural networks for prediction, they report errors (mean ± SD, in number of slices) of, respectively, 2.7 ± 1.8, 3.0 ± 3.0, 3.7 ± 1.9, 3.9 ± 3.2, and 2.5 ± 1.8 for these five body region boundaries.…”

“…An important step toward effective implementation of quantitative radiology is the recognition and delineation of objects in the human body. An essential part of this task is standardizing definitions for such objects, which may be organs, tissue regions, or well-defined body regions, 1,2 in order to develop generalizable body-wide methods, standardize clinical operations, and use the quantitative information meaningfully. With this ideology in mind, the standardization of body region boundaries is an equally important task, especially for objects that span body regions such as the thoracic spinal cord, the boundaries of which clearly depend on the definition of the thoracic region.…”

“…Bai et al 1 tackle our exact problem by using a system of virtual landmarks employing principal component analysis and recursive subdivision of objects, and subsequently using a neural network for mapping the virtual landmarks to boundary locations. Hussein et al 4 propose a one‐shot deep learning solution for automatically localizing the boundaries for the abdominal and thoracic regions by locating the superior and inferior boundaries of the abdomen and thorax, respectively, as a step in segmenting and quantifying intrathoracic adipose tissue in positron emission tomography/CT (PET/CT) images.…”

“…Our system BRR‐Net is designed to perform all three output actions O1‐O3 and output the following: Superior boundary of H&N: slice 10; inferior boundary of H&N: slice 60; H&N body region: slices 10‐60; superior boundary of thorax: slice 70; inferior boundary of thorax: not in I ; thorax body region: incomplete. Bai et al 1 focus on three body regions – thorax, abdomen, and pelvis, and do not consider H&N, although it is possible to extend their method to H&N. Hussein et al 4 consider only the thorax and abdomen. More important than the smaller number of body regions considered by both works, their behavior becomes unpredictable (or not described/demonstrated in the papers) if I does not include the slices they aim to detect.…”