skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: TU-H-CAMPUS-JeP2-02: Interobserver Variability of CT, PET-CT and MRI Based Primary Tumor Delineation for Lung Cancer

Abstract

Purpose: Target delineation in lung cancer radiotherapy has, in general, large variability. MRI has so far not been investigated in detail for lung cancer delineation variability. The purpose of this study is to investigate delineation variability for lung tumors using MRI and compare it to CT alone and PET-CT based delineations. Methods: Seven physicians delineated the primary tumor volumes of nine patients for the following scenarios: (1) CT only; (2) post-contrast T1-weighted MRI registered with diffusion-weighted MRI; and (3) PET-CT fusion images. To compute interobserver variability, the median surface was generated from all observers’ contours and used as the reference surface. A single physician labeled the interface types (tumor to lung, atelectasis (collapsed lung), hilum, mediastinum, or chest-wall) on the median surface. Volume variation (normalized to PET-CT volume), minimum distance (MD), and bidirectional local distance (BLD) between individual observers’ contours and the reference contour were measured. Results: CT- and MRI-based normalized volumes were 1.61±0.76 (mean±SD) and 1.38±0.44, respectively, both significantly larger than PET-CT (p<0.05, paired t-test). The overall uncertainty (root mean square of SD values over all points) of both BLD and MD measures of the observers for the interfaces were not significantly different (p>0.05, two-samples t-test) for all imagingmore » modalities except between tumor-mediastinum and tumor-atelectasis in PET-CT. The largest mean overall uncertainty was observed for tumor-atelectasis interface, the smallest for tumor-mediastinum and tumor-lung interfaces for all modalities. The whole tumor uncertainties for both BLD and MD were not significantly different between any two modalities (p>0.05, paired t-test). Overall uncertainties for the interfaces using BLD were similar to using MD. Conclusion: Large volume variations were observed between the three imaging modalities. Contouring variability appeared to depend on the interface type. This study will be useful for understanding the delineation uncertainty for radiotherapy planning of lung cancer using different imaging modalities. Disclosures: Research agreement with Phillips Healthcare (GH and EW), National Institutes of Health Licensing agreement with Varian Medical Systems (GH and EW), research grants from the National Institute of Health (GH and EW), UpToDate royalties (EW), and none (others). Authors have no potential conflicts of interest to disclose.« less

Authors:
; ; ; ; ; ; ; ; ;  [1]
  1. Virginia Commonwealth University, Richmond, VA (United States)
Publication Date:
OSTI Identifier:
22654061
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; BIOMEDICAL RADIOGRAPHY; IMAGES; INTERFACES; LUNGS; NEOPLASMS; NMR IMAGING; POSITRON COMPUTED TOMOGRAPHY

Citation Formats

Karki, K, Hugo, G, Saraiya, S, Jan, N, Schuster, J, Schutzer, M, Fahrner, L, Groves, R, Ford, J, and Weiss, E. TU-H-CAMPUS-JeP2-02: Interobserver Variability of CT, PET-CT and MRI Based Primary Tumor Delineation for Lung Cancer. United States: N. p., 2016. Web. doi:10.1118/1.4957685.
Karki, K, Hugo, G, Saraiya, S, Jan, N, Schuster, J, Schutzer, M, Fahrner, L, Groves, R, Ford, J, & Weiss, E. TU-H-CAMPUS-JeP2-02: Interobserver Variability of CT, PET-CT and MRI Based Primary Tumor Delineation for Lung Cancer. United States. doi:10.1118/1.4957685.
Karki, K, Hugo, G, Saraiya, S, Jan, N, Schuster, J, Schutzer, M, Fahrner, L, Groves, R, Ford, J, and Weiss, E. 2016. "TU-H-CAMPUS-JeP2-02: Interobserver Variability of CT, PET-CT and MRI Based Primary Tumor Delineation for Lung Cancer". United States. doi:10.1118/1.4957685.
@article{osti_22654061,
title = {TU-H-CAMPUS-JeP2-02: Interobserver Variability of CT, PET-CT and MRI Based Primary Tumor Delineation for Lung Cancer},
author = {Karki, K and Hugo, G and Saraiya, S and Jan, N and Schuster, J and Schutzer, M and Fahrner, L and Groves, R and Ford, J and Weiss, E},
abstractNote = {Purpose: Target delineation in lung cancer radiotherapy has, in general, large variability. MRI has so far not been investigated in detail for lung cancer delineation variability. The purpose of this study is to investigate delineation variability for lung tumors using MRI and compare it to CT alone and PET-CT based delineations. Methods: Seven physicians delineated the primary tumor volumes of nine patients for the following scenarios: (1) CT only; (2) post-contrast T1-weighted MRI registered with diffusion-weighted MRI; and (3) PET-CT fusion images. To compute interobserver variability, the median surface was generated from all observers’ contours and used as the reference surface. A single physician labeled the interface types (tumor to lung, atelectasis (collapsed lung), hilum, mediastinum, or chest-wall) on the median surface. Volume variation (normalized to PET-CT volume), minimum distance (MD), and bidirectional local distance (BLD) between individual observers’ contours and the reference contour were measured. Results: CT- and MRI-based normalized volumes were 1.61±0.76 (mean±SD) and 1.38±0.44, respectively, both significantly larger than PET-CT (p<0.05, paired t-test). The overall uncertainty (root mean square of SD values over all points) of both BLD and MD measures of the observers for the interfaces were not significantly different (p>0.05, two-samples t-test) for all imaging modalities except between tumor-mediastinum and tumor-atelectasis in PET-CT. The largest mean overall uncertainty was observed for tumor-atelectasis interface, the smallest for tumor-mediastinum and tumor-lung interfaces for all modalities. The whole tumor uncertainties for both BLD and MD were not significantly different between any two modalities (p>0.05, paired t-test). Overall uncertainties for the interfaces using BLD were similar to using MD. Conclusion: Large volume variations were observed between the three imaging modalities. Contouring variability appeared to depend on the interface type. This study will be useful for understanding the delineation uncertainty for radiotherapy planning of lung cancer using different imaging modalities. Disclosures: Research agreement with Phillips Healthcare (GH and EW), National Institutes of Health Licensing agreement with Varian Medical Systems (GH and EW), research grants from the National Institute of Health (GH and EW), UpToDate royalties (EW), and none (others). Authors have no potential conflicts of interest to disclose.},
doi = {10.1118/1.4957685},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • Purpose: To compare source-to-background ratio (SBR)-based PET-CT auto-delineation with pathology in non-small-cell lung cancer (NSCLC) and to investigate whether auto-delineation reduces the interobserver variability compared with manual PET-CT-based gross tumor volume (GTV) delineation. Methods and Materials: Source-to-background ratio-based auto-delineation was compared with macroscopic tumor dimensions to assess its validity in 23 tumors. Thereafter, GTVs were delineated manually on 33 PET-CT scans by five observers for the primary tumor (GTV-1) and the involved lymph nodes (GTV-2). The delineation was repeated after 6 months with the auto-contour provided. This contour was edited by the observers. For comparison, the concordance index (CI) wasmore » calculated, defined as the ratio of intersection and the union of two volumes (A intersection B)/(A union B). Results: The maximal tumor diameter of the SBR-based auto-contour correlated strongly with the macroscopic diameter of primary tumors (correlation coefficient = 0.90) and was shown to be accurate for involved lymph nodes (sensitivity 67%, specificity 95%). The median auto-contour-based target volumes were smaller than those defined by manual delineation for GTV-1 (31.8 and 34.6 cm{sup 3}, respectively; p = 0.001) and GTV-2 (16.3 and 21.8 cm{sup 3}, respectively; p 0.02). The auto-contour-based method showed higher CIs than the manual method for GTV-1 (0.74 and 0.70 cm{sup 3}, respectively; p < 0.001) and GTV-2 (0.60 and 0.51 cm{sup 3}, respectively; p = 0.11). Conclusion: Source-to-background ratio-based auto-delineation showed a good correlation with pathology, decreased the delineated volumes of the GTVs, and reduced the interobserver variability. Auto-contouring may further improve the quality of target delineation in NSCLC patients.« less
  • Purpose: In radiation treatment planning, delineation of gross tumor volume (GTV) is very important, because the GTVs affect the accuracies of radiation therapy procedure. To assist radiation oncologists in the delineation of GTV regions while treatment planning for lung cancer, we have proposed a machine-learning-based delineation framework of GTV regions of solid and ground glass opacity (GGO) lung tumors following by optimum contour selection (OCS) method. Methods: Our basic idea was to feed voxel-based image features around GTV contours determined by radiation oncologists into a machine learning classifier in the training step, after which the classifier produced the degree ofmore » GTV for each voxel in the testing step. Ten data sets of planning CT and PET/CT images were selected for this study. The support vector machine (SVM), which learned voxel-based features which include voxel value and magnitudes of image gradient vector that obtained from each voxel in the planning CT and PET/CT images, extracted initial GTV regions. The final GTV regions were determined using the OCS method that was able to select a global optimum object contour based on multiple active delineations with a level set method around the GTV. To evaluate the results of proposed framework for ten cases (solid:6, GGO:4), we used the three-dimensional Dice similarity coefficient (DSC), which denoted the degree of region similarity between the GTVs delineated by radiation oncologists and the proposed framework. Results: The proposed method achieved an average three-dimensional DSC of 0.81 for ten lung cancer patients, while a standardized uptake value-based method segmented GTV regions with the DSC of 0.43. The average DSCs for solid and GGO were 0.84 and 0.76, respectively, obtained by the proposed framework. Conclusion: The proposed framework with the support vector machine may be useful for assisting radiation oncologists in delineating solid and GGO lung tumors.« less
  • Purpose: Breathing consistency variations can cause respiratory-related motion blurring and artifacts and increase in MRI scan time due to inadequate respiratory-gating and discarding of breathing cycles. In a previous study the concept of audiovisual biofeedback (AV) guided respiratory-gated MRI was tested with healthy volunteers and it demonstrated image quality improvement on anatomical structures and scan time reduction. This study tests the applicability of AV-guided respiratorygated MRI for lung cancer in a prospective patient study. Methods: Image quality and scan time were investigated in thirteen lung cancer patients who underwent two 3T MRI sessions. In the first MRI session (pre-treatment), respiratory-gatedmore » MR images with free breathing (FB) and AV were acquired at inhalation and exhalation. An RF navigator placed on the liver dome was employed for the respiratory-gated MRI. This was repeated in the second MRI session (mid-treatment). Lung tumors were delineated on each dataset. FB and AV were compared in terms of (1) tumor definition assessed by lung tumor contours and (2) intra-patient scan time variation using the total image acquisition time of inhalation and exhalation datasets from the first and second MRI sessions across 13 lung cancer patients. Results: Compared to FB AV-guided respiratory-gated MRI improved image quality for contouring tumors with sharper boundaries and less blurring resulted in the improvement of tumor definition. Compared to FB the variation of intra-patient scan time with AV was reduced by 48% (p<0.001) from 54 s to 28 s. Conclusion: This study demonstrated that AV-guided respiratorygated MRI improved the quality of tumor images and fixed tumor definition for lung cancer. These results suggest that audiovisual biofeedback breathing guidance has the potential to control breathing for adequate respiratory-gating for lung cancer imaging and radiotherapy.« less
  • Purpose: It is generally agreed that the safe implementation of stereotactic body radiotherapy requires image guidance. The aim of this work was to assess interobserver variability in the delineation of lung lesions on cone-beam CT (CBCT) images compared with CT-based contouring for adaptive stereotactic body radiotherapy. The influence of target size was also evaluated. Methods and Materials: Eight radiation oncologists delineated gross tumor volumes in 12 patient cases (non-small cell lung cancer I-II or solitary metastasis) on planning CTs and on CBCTs. Cases were divided into two groups with tumor diameters of less than (Group A) or more than 2more » cm (Group B). Comparison of mean volumes delineated by all observers and range and coefficient of variation were reported for each case and image modality. Interobserver variability was assessed by means of standard error of measurement, conformity index (CI), and its generalized observer-independent approach. The variance between single observers on CT and CBCT images was measured via interobserver reliability coefficient. Results: Interobserver variability on CT images was 17% with 0.79 reliability, compared with 21% variability on CBCT and 0.76 reliability. On both image modalities, values of the intraobserver reliability coefficient (0.99 for CT and 0.97 for CBCT) indicated high reproducibility of results. In general, lower interobserver agreement was observed for small lesions (CI{sub genA} = 0.62 {+-} 0.06 vs. CI{sub genB} = 0.70 {+-} 0.03, p < 0.05). The analysis of single patient cases revealed that presence of spicules, diffuse infiltrations, proximity of the tumors to the vessels and thoracic wall, and respiration motion artifacts presented the main sources of the variability. Conclusion: Interobserver variability for Stage I-II non-small cell lung cancer and lung metastasis was slightly higher on CBCT compared with CT. Absence of significant differences in interobserver variability suggests that CBCT imaging provides an effective tool for tumor localization, and image data could be also used for target volume delineation purposes.« less
  • Purpose: Positron emission tomography (PET) with the glucose analogue [18F] fluoro-2-deoxy-D-glucose ({sup 18}F-FDG-PET) has been used in radiation treatment planning for non-small-cell carcinoma. To date, lymph nodes have been contoured according to the uptake of the tumor. This prospective study was performed to evaluate if nodal volume delineates according to FDG uptake within the primary tumor (PET-GTVnt) is suitable for nodal target volume delineation or if individualized nodal FDG uptake measure (PET-GTVnn) is necessary to better nodal target definition. Methods and Materials: Forty cases, who underwent a diagnostic {sup 18}F-FDG PET/computed tomography (CT) scan, were included. Two PET-based GTVs formore » each lymph node were contoured and compared. First, we used an isocontour of 40% of the maximum tumor uptake (PET-GTVnt). Second, an isocontour of 40% of the maximum uptake of each node (PET-GTVnn) was employed. To avoid interobserver variability, this was carried out by the same radiation oncologist. Afterwards, the difference between both lymph node volumes was plotted against the ratio of the maximum uptakes (I{sub n}/I{sub t}) in a linear regression analysis. Results: Compared with CT-based lymph node volume (CT-GTVn), the intraclass correlation coefficient of PET-GTVnn was higher than the coefficient of PET-GTVnt (p < 0.001). All cases could be divided into four groups: undetected (17.5%), detected but overestimated (10%), detected but underestimated (35%), and correctly detected (37.5%). Conclusions: If a method of automatic delineation shall be applied, this method must be applied to every lesion separately. However, to facilitate the delineation in daily practice, when I{sub n}/I{sub t} is {<=}25%, lymph nodes could be delineated in accordance with tumor uptake, keeping an absolute difference in radii <5 mm.« less