3
|
Vorontsov E, Cerny M, Régnier P, Di Jorio L, Pal CJ, Lapointe R, Vandenbroucke-Menu F, Turcotte S, Kadoury S, Tang A. Deep Learning for Automated Segmentation of Liver Lesions at CT in Patients with Colorectal Cancer Liver Metastases. Radiol Artif Intell 2019; 1:180014. [PMID: 33937787 DOI: 10.1148/ryai.2019180014] [Citation(s) in RCA: 52] [Impact Index Per Article: 10.4] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/28/2018] [Revised: 01/25/2019] [Accepted: 01/31/2019] [Indexed: 02/06/2023]
Abstract
Purpose To evaluate the performance, agreement, and efficiency of a fully convolutional network (FCN) for liver lesion detection and segmentation at CT examinations in patients with colorectal liver metastases (CLMs). Materials and Methods This retrospective study evaluated an automated method using an FCN that was trained, validated, and tested with 115, 15, and 26 contrast material-enhanced CT examinations containing 261, 22, and 105 lesions, respectively. Manual detection and segmentation by a radiologist was the reference standard. Performance of fully automated and user-corrected segmentations was compared with that of manual segmentations. The interuser agreement and interaction time of manual and user-corrected segmentations were assessed. Analyses included sensitivity and positive predictive value of detection, segmentation accuracy, Cohen κ, Bland-Altman analyses, and analysis of variance. Results In the test cohort, for lesion size smaller than 10 mm (n = 30), 10-20 mm (n = 35), and larger than 20 mm (n = 40), the detection sensitivity of the automated method was 10%, 71%, and 85%; positive predictive value was 25%, 83%, and 94%; Dice similarity coefficient was 0.14, 0.53, and 0.68; maximum symmetric surface distance was 5.2, 6.0, and 10.4 mm; and average symmetric surface distance was 2.7, 1.7, and 2.8 mm, respectively. For manual and user-corrected segmentation, κ values were 0.42 (95% confidence interval: 0.24, 0.63) and 0.52 (95% confidence interval: 0.36, 0.72); normalized interreader agreement for lesion volume was -0.10 ± 0.07 (95% confidence interval) and -0.10 ± 0.08; and mean interaction time was 7.7 minutes ± 2.4 (standard deviation) and 4.8 minutes ± 2.1 (P < .001), respectively. Conclusion Automated detection and segmentation of CLM by using deep learning with convolutional neural networks, when manually corrected, improved efficiency but did not substantially change agreement on volumetric measurements.© RSNA, 2019Supplemental material is available for this article.
Collapse
Affiliation(s)
- Eugene Vorontsov
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Milena Cerny
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Philippe Régnier
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Lisa Di Jorio
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Christopher J Pal
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Réal Lapointe
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Franck Vandenbroucke-Menu
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Simon Turcotte
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - Samuel Kadoury
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| | - An Tang
- Department of Radiology (M.C., A.T.) and Department of Surgery, Hepatopancreatobiliary and Liver Transplantation Division (R.L., F.V., S.T.), Centre Hospitalier de l'Université de Montréal (CHUM), 1000 rue Saint-Denis, Montréal, QC, Canada H2X 0C2; Montreal Institute for Learning Algorithms (MILA), Montréal, Canada (E.V., C.J.P.); École Polytechnique, Montréal, Canada (E.V., C.J.P., S.K.); Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada (M.C., P.R., S.T., S.K., A.T.); and Imagia Cybernetics, Montréal, Canada (L.D.J.)
| |
Collapse
|
7
|
Helou J, Karotki A, Milot L, Chu W, Erler D, Chung HT. 4DCT Simulation With Synchronized Contrast Injection in Liver SBRT Patients. Technol Cancer Res Treat 2015; 15:55-9. [PMID: 25731803 DOI: 10.1177/1533034615572341] [Citation(s) in RCA: 11] [Impact Index Per Article: 1.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/07/2014] [Accepted: 01/20/2015] [Indexed: 12/26/2022] Open
Abstract
BACKGROUND/PURPOSE Delivering stereotactic body radiotherapy for liver metastases remains a challenge because of respiratory motion and poor visibility without intravenous contrast. The purpose of this article is to describe a novel and simple computed tomography (CT) simulation process of integrating timed intravenous contrast that could overcome the uncertainty of target delineation. METHODS AND RESULTS The simulation involves two 4-dimensional CT (4DCT) scans. The first scan only encompasses the immediate region of the tumor and surrounding tissue, which reduces the 4DCT scan time so that it can be optimally timed with intravenous contrast injection. The second 4DCT scan covers a larger volume and is used as the primary CT data set for dose calculation, as well as patient setup verification on the treatment unit. The combination of the two 4DCT scans allows us to optimally visualize liver metastases over all phases of the breathing cycle while simultaneously acquiring a long enough 4DCT data set that is suitable for planning and patient setup verification. CONCLUSION This simulation technique allows for a better target definition when treating liver metastases, without being invasive.
Collapse
Affiliation(s)
- Joelle Helou
- Department of Radiation Oncology, Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
| | - Aliaksandr Karotki
- Department of Medical Physics, Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada
| | - Laurent Milot
- Department of Medical Imaging, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada
| | - William Chu
- Department of Radiation Oncology, Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
| | - Darby Erler
- Department of Radiation Oncology, Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada
| | - Hans T Chung
- Department of Radiation Oncology, Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto, Ontario, Canada Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
| |
Collapse
|
8
|
Albrecht MH, Wichmann JL, Müller C, Schreckenbach T, Sakthibalan S, Hammerstingl R, Bechstein WO, Zangos S, Ackermann H, Vogl TJ. Assessment of colorectal liver metastases using MRI and CT: impact of observer experience on diagnostic performance and inter-observer reproducibility with histopathological correlation. Eur J Radiol 2014; 83:1752-8. [PMID: 25082480 DOI: 10.1016/j.ejrad.2014.07.005] [Citation(s) in RCA: 9] [Impact Index Per Article: 0.9] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/04/2014] [Revised: 07/01/2014] [Accepted: 07/05/2014] [Indexed: 02/07/2023]
Abstract
INTRODUCTION To compare the diagnostic performance and inter-observer reproducibility of CT and MRI in detecting colorectal liver metastases (CRLM) of observers with different levels of experience. MATERIALS AND METHODS Data from 51 CT and 54 MRI examinations of 105 patients with CRLM were analysed. Intraoperative and histopathological findings served as the reference standard. Analyses were performed by four observers with varying levels of experience regarding imaging of CRLM (reviewers A, B, C and D with respectively >20, >5, <1 and 0 years of experience). Per-segment sensitivity, specificity, Cohen's kappa (κ) for diagnosed segments and Intra-class Correlation Coefficients (ICC) for reported number of lesions were calculated. RESULTS CT sensitivity and specificity was for reviewer A 89.71%/94.41%, B 78.50%/88.37%, C 63.55%/85.58%, D 84.11%/78.60% and regarding MRI A 90.40%/95.43%, B 74.40%/90.04%, C 60.00%/85.89% and D 65.60%/75.90%. The overall inter-observer agreement was higher for CT (κ=0.43, p<0.001; ICC=0.75, p<0.001) than MRI (κ=0.38, p<0.001; ICC=0.65, p<0.001). The experienced reviewers A and B achieved better agreement for MRI (κ=0.54, p<0.001; ICC=0.77, p<0.001) than CT (κ=0.52, p<0.00; ICC=0.76, p<0.001) unlike the less experienced C and D (MRI κ=0.38, ICC=0.63 and CT κ=0.41, ICC=0.74, respectively, p<0.001). CONCLUSIONS The proficiency in detection of CRLM is significantly influenced by observer experience, although CT interpretation is less affected than MRI analysis.
Collapse
Affiliation(s)
- Moritz H Albrecht
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany.
| | - Julian L Wichmann
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Cindy Müller
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Theresa Schreckenbach
- University Hospital Frankfurt, Department of General and Visceral Surgery, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Sreekanth Sakthibalan
- Barts and the London, Queen Mary University of London, Mile End Road, London E1 4NS, United Kingdom
| | - Renate Hammerstingl
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Wolf O Bechstein
- University Hospital Frankfurt, Department of General and Visceral Surgery, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Stephan Zangos
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Hanns Ackermann
- University Hospital Frankfurt, Department of Biostatistics and Medical Information, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| | - Thomas J Vogl
- University Hospital Frankfurt, Department of Diagnostic and Interventional Radiology, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany
| |
Collapse
|