Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models
Radiomics investigates the predictive role of quantitative parameters calculated from radiological images. In oncology, tumour segmentation constitutes a crucial step of the radiomic workflow. Manual segmentation is time-consuming and prone to inter-observer variability. In this study, a state-of-th...
Main Authors: | , , , , , , , , , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2022-12-01
|
Series: | Journal of Clinical Medicine |
Subjects: | |
Online Access: | https://www.mdpi.com/2077-0383/11/24/7334 |
_version_ | 1797457102328823808 |
---|---|
author | Matteo Ferrante Lisa Rinaldi Francesca Botta Xiaobin Hu Andreas Dolp Marta Minotti Francesca De Piano Gianluigi Funicelli Stefania Volpe Federica Bellerba Paolo De Marco Sara Raimondi Stefania Rizzo Kuangyu Shi Marta Cremonesi Barbara A. Jereczek-Fossa Lorenzo Spaggiari Filippo De Marinis Roberto Orecchia Daniela Origgi |
author_facet | Matteo Ferrante Lisa Rinaldi Francesca Botta Xiaobin Hu Andreas Dolp Marta Minotti Francesca De Piano Gianluigi Funicelli Stefania Volpe Federica Bellerba Paolo De Marco Sara Raimondi Stefania Rizzo Kuangyu Shi Marta Cremonesi Barbara A. Jereczek-Fossa Lorenzo Spaggiari Filippo De Marinis Roberto Orecchia Daniela Origgi |
author_sort | Matteo Ferrante |
collection | DOAJ |
description | Radiomics investigates the predictive role of quantitative parameters calculated from radiological images. In oncology, tumour segmentation constitutes a crucial step of the radiomic workflow. Manual segmentation is time-consuming and prone to inter-observer variability. In this study, a state-of-the-art deep-learning network for automatic segmentation (nnU-Net) was applied to computed tomography images of lung tumour patients, and its impact on the performance of survival radiomic models was assessed. In total, 899 patients were included, from two proprietary and one public datasets. Different network architectures (2D, 3D) were trained and tested on different combinations of the datasets. Automatic segmentations were compared to reference manual segmentations performed by physicians using the DICE similarity coefficient. Subsequently, the accuracy of radiomic models for survival classification based on either manual or automatic segmentations were compared, considering both hand-crafted and deep-learning features. The best agreement between automatic and manual contours (DICE = 0.78 ± 0.12) was achieved averaging 2D and 3D predictions and applying customised post-processing. The accuracy of the survival classifier (ranging between 0.65 and 0.78) was not statistically different when using manual versus automatic contours, both with hand-crafted and deep features. These results support the promising role nnU-Net can play in automatic segmentation, accelerating the radiomic workflow without impairing the models’ accuracy. Further investigations on different clinical endpoints and populations are encouraged to confirm and generalise these findings. |
first_indexed | 2024-03-09T16:17:21Z |
format | Article |
id | doaj.art-fb2ebf3e00534ce9a092c2d8995b4c9f |
institution | Directory Open Access Journal |
issn | 2077-0383 |
language | English |
last_indexed | 2024-03-09T16:17:21Z |
publishDate | 2022-12-01 |
publisher | MDPI AG |
record_format | Article |
series | Journal of Clinical Medicine |
spelling | doaj.art-fb2ebf3e00534ce9a092c2d8995b4c9f2023-11-24T15:43:47ZengMDPI AGJournal of Clinical Medicine2077-03832022-12-011124733410.3390/jcm11247334Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic ModelsMatteo Ferrante0Lisa Rinaldi1Francesca Botta2Xiaobin Hu3Andreas Dolp4Marta Minotti5Francesca De Piano6Gianluigi Funicelli7Stefania Volpe8Federica Bellerba9Paolo De Marco10Sara Raimondi11Stefania Rizzo12Kuangyu Shi13Marta Cremonesi14Barbara A. Jereczek-Fossa15Lorenzo Spaggiari16Filippo De Marinis17Roberto Orecchia18Daniela Origgi19Medical Physics Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyRadiation Research Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyMedical Physics Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDepartment of Informatics, Technical University of Munich, Arcisstraße 21, 80333 Munich, GermanyDepartment of Informatics, Technical University of Munich, Arcisstraße 21, 80333 Munich, GermanyDivision of Radiology, IEO, European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDivision of Radiology, IEO, European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDivision of Radiology, IEO, European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDivision of Radiation Oncology, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDepartment of Experimental Oncology, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyMedical Physics Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDepartment of Experimental Oncology, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyClinica di Radiologia EOC, Istituto Imaging della Svizzera Italiana (IIMSI), via Tesserete 46, 6900 Lugano, SwitzerlandChair for Computer-Aided Medical Procedures, Department of Informatics, Technical University of Munich, Arcisstraße 21, 80333 Munich, GermanyRadiation Research Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDivision of Radiation Oncology, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDepartment of Oncology and Hemato-Oncology, University of Milan, via Festa del Perdono 7, 20122 Milan, ItalyDivision of Thoracic Oncology, IEO, European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyDivision of Radiology, IEO, European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyMedical Physics Unit, IEO European Institute of Oncology IRCCS, via Ripamonti 435, 20141 Milan, ItalyRadiomics investigates the predictive role of quantitative parameters calculated from radiological images. In oncology, tumour segmentation constitutes a crucial step of the radiomic workflow. Manual segmentation is time-consuming and prone to inter-observer variability. In this study, a state-of-the-art deep-learning network for automatic segmentation (nnU-Net) was applied to computed tomography images of lung tumour patients, and its impact on the performance of survival radiomic models was assessed. In total, 899 patients were included, from two proprietary and one public datasets. Different network architectures (2D, 3D) were trained and tested on different combinations of the datasets. Automatic segmentations were compared to reference manual segmentations performed by physicians using the DICE similarity coefficient. Subsequently, the accuracy of radiomic models for survival classification based on either manual or automatic segmentations were compared, considering both hand-crafted and deep-learning features. The best agreement between automatic and manual contours (DICE = 0.78 ± 0.12) was achieved averaging 2D and 3D predictions and applying customised post-processing. The accuracy of the survival classifier (ranging between 0.65 and 0.78) was not statistically different when using manual versus automatic contours, both with hand-crafted and deep features. These results support the promising role nnU-Net can play in automatic segmentation, accelerating the radiomic workflow without impairing the models’ accuracy. Further investigations on different clinical endpoints and populations are encouraged to confirm and generalise these findings.https://www.mdpi.com/2077-0383/11/24/7334nnU-NetNSCLCautomatic segmentationradiomicshand-crafted/deep featurespredictive model |
spellingShingle | Matteo Ferrante Lisa Rinaldi Francesca Botta Xiaobin Hu Andreas Dolp Marta Minotti Francesca De Piano Gianluigi Funicelli Stefania Volpe Federica Bellerba Paolo De Marco Sara Raimondi Stefania Rizzo Kuangyu Shi Marta Cremonesi Barbara A. Jereczek-Fossa Lorenzo Spaggiari Filippo De Marinis Roberto Orecchia Daniela Origgi Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models Journal of Clinical Medicine nnU-Net NSCLC automatic segmentation radiomics hand-crafted/deep features predictive model |
title | Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models |
title_full | Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models |
title_fullStr | Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models |
title_full_unstemmed | Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models |
title_short | Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models |
title_sort | application of nnu net for automatic segmentation of lung lesions on ct images and its implication for radiomic models |
topic | nnU-Net NSCLC automatic segmentation radiomics hand-crafted/deep features predictive model |
url | https://www.mdpi.com/2077-0383/11/24/7334 |
work_keys_str_mv | AT matteoferrante applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT lisarinaldi applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT francescabotta applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT xiaobinhu applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT andreasdolp applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT martaminotti applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT francescadepiano applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT gianluigifunicelli applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT stefaniavolpe applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT federicabellerba applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT paolodemarco applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT sararaimondi applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT stefaniarizzo applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT kuangyushi applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT martacremonesi applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT barbaraajereczekfossa applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT lorenzospaggiari applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT filippodemarinis applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT robertoorecchia applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels AT danielaoriggi applicationofnnunetforautomaticsegmentationoflunglesionsonctimagesanditsimplicationforradiomicmodels |