Pulmonary nodule registration: Rigid or nonrigid?

Abstract

Purpose: The primary aim of this study is to investigate the performance difference of rigid and nonrigid registration schemes in matching corresponding pulmonary nodules depicted on sequential chest computed tomography (CT) examinations. Methods: A gradient descent based rigid registration algorithm with scaling was developed and it handled the involved geometric transformations (i.e., translation, rescaling, shearing, and rotation) separately instead of optimizing them in a single pass. Given two lung CT examinations, the scaling and translation parameters were simply estimated from the lung volume dimensions (e.g., size and mass center), while the rotation parameters were optimized progressively using gradient descent. To investigate the performance difference of rigid and nonrigid schemes in pulmonary nodule registration, the well-known nonrigid Demons algorithm was implemented and tested along with the developed schemes against 60 diverse low-dose clinical lung CT examinations with average 2-yr follow-up scans. A verified cancer and its correspondence in the follow-up scan as well as their spatial locations (mass center) were identified in each examination. In addition to the computational efficiency, the accuracy of these registration procedures was assessed by computing the Euclidean distances between the corresponding nodules after the registration. To demonstrate the advantage of the developed algorithm, the authors also implemented a fast iterative closest point (ICP) based rigid algorithm and compared their performance. Results: Our experiments on the collected chest CT examinations showed that the nodule registration errors in 3D Euclidean distance for the developed rigid affine approach, the traditional ICP algorithm, and the refining nonrigid Demons algorithm were 9.6, 9.8, and 10.0 mm, respectively, and the corresponding computational costs in time were 5, 300, and 55 s, respectively. Conclusions: A rigid solution may be preferred in practice for the pulmonary nodule registration in longitudinal studies because of its relatively high efficiency and sufficient accuracy for the clinical need.

Correlation of normal lung density changes with dose after stereotactic body radiotherapy (SBRT) for early stage lung cancer

“…However, it should be noted that it is quite challenging to perform deformable registration of the initial tumor volume on planning CT to a significantly different or even nonexistent tumor volume on the follow-up CT. Observed differences in lung density were, in part, likely due to suboptimal coregistration of planning and follow-up CT scans, although rigid registration was shown to be a simpler, yet valid approach in a study by Gu et al in the context of follow-up examinations for lung nodules [64].…”

Correlation of Clinical and Dosimetric Parameters With Radiographic Lung Injury Following Stereotactic Body Radiotherapy

“…Some investigators have argued that, due to respiratory motion causing either inflation or deflation of the lung parenchyma, DR should be utilized when comparing CT scans. 38,39 However, in a series examining different registration algorithms for 60 pulmonary nodules compared to follow-up scans at an average of 2 years followup, Gu et al showed that rigid registration (RR) had equivalent accuracy compared with DR. 40 In their series, Diot et al found discrepancies of <5% for doses less than 60 Gy in a subset of their patients when comparing DR and RR. 17 We performed a subset analysis using DR in 23 patients for geometric modeling purposes.…”

Selecting registration schemes in case of interstitial lung disease follow‐up in CT

“…Although nonrigid registration schemes seem the natural choice for lung analysis, due to the elastic nature of lung tissue, combination schemes that comprised of rigid and nonrigid transforms are recently proposed, considering the trade-off between sufficient accuracy and time efficiency in the clinical environment. 12 Registration schemes evaluated in the present study were initially selected on the basis of artificially warped ground truth follow-up data, generated from original baseline images of ten clinical volumetric CT scans, in order to avoid T IV. Average distance error (mean ± standard deviation) of corresponding landmark points in normal lung parenchyma regions and interstitial lung disease affected regions between original baseline and follow-up data registered by the selected registration schemes, as well as between original baseline and unregistered follow-up data.…”

“…Although nonrigid transformation models seem the natural choice for lung field registration due to the elastic nature of the lung tissue, combinations of rigid and nonrigid transformation models were also included, as recently reported in the lung CT registration literature. 12 The B-spline grid-spacing applied was 3 × 3 × 3 voxels (approximately 13.2 cm 3 ). Finally, four different gradient descent optimizers were considered (standard gradient decent-SGD, regular step gradient decent-RSGD, adaptive stochastic gradient decent-ASGD, and finite difference gradient decent-FDGD).…”

“…Up to now, only one study has reported on ILD follow-up monitoring in the context of 2D axial slices of baseline and follow-up scans, subjected to volume registration. 7 Besides ILD monitoring, image registration has received considerable attention in lung CT image analysis applications, such as atlas registration-based segmentation, 8 lung nodule monitoring, [9][10][11][12] emphysema monitoring and estimation of local ventilation and perfusion, 13,14 and motion estimation for radiotherapy planning, 15,16 exploiting various inspiration/expiration protocols. [17][18][19] Multiresolution nonrigid registration, capable for capturing local lung tissue deformations, accounts for a commonly used approach for lung CT registration utilized in all resolution levels, or in combination with rigid transforms applied in low resolution levels for computational time considerations.…”