Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography
Background and purpose: To improve cone-beam computed tomography (CBCT), deep-learning (DL)-models are being explored to generate synthetic CTs (sCT). The sCT evaluation is mainly focused on image quality and CT number accuracy. However, correct representation of daily anatomy of the CBCT is also im...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2023-01-01
|
Series: | Physics and Imaging in Radiation Oncology |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2405631623000076 |
_version_ | 1797870499064184832 |
---|---|
author | Yvonne J.M. de Hond Camiel E.M. Kerckhaert Maureen A.J.M. van Eijnatten Paul M.A. van Haaren Coen W. Hurkmans Rob H.N. Tijssen |
author_facet | Yvonne J.M. de Hond Camiel E.M. Kerckhaert Maureen A.J.M. van Eijnatten Paul M.A. van Haaren Coen W. Hurkmans Rob H.N. Tijssen |
author_sort | Yvonne J.M. de Hond |
collection | DOAJ |
description | Background and purpose: To improve cone-beam computed tomography (CBCT), deep-learning (DL)-models are being explored to generate synthetic CTs (sCT). The sCT evaluation is mainly focused on image quality and CT number accuracy. However, correct representation of daily anatomy of the CBCT is also important for sCTs in adaptive radiotherapy. The aim of this study was to emphasize the importance of anatomical correctness by quantitatively assessing sCT scans generated from CBCT scans using different paired and unpaired dl-models. Materials and methods: Planning CTs (pCT) and CBCTs of 56 prostate cancer patients were included to generate sCTs. Three different dl-models, Dual-UNet, Single-UNet and Cycle-consistent Generative Adversarial Network (CycleGAN), were evaluated on image quality and anatomical correctness. The image quality was assessed using image metrics, such as Mean Absolute Error (MAE). The anatomical correctness between sCT and CBCT was quantified using organs-at-risk volumes and average surface distances (ASD). Results: MAE was 24 Hounsfield Unit (HU) [range:19-30 HU] for Dual-UNet, 40 HU [range:34-56 HU] for Single-UNet and 41HU [range:37-46 HU] for CycleGAN. Bladder ASD was 4.5 mm [range:1.6–12.3 mm] for Dual-UNet, 0.7 mm [range:0.4–1.2 mm] for Single-UNet and 0.9 mm [range:0.4–1.1 mm] CycleGAN. Conclusions: Although Dual-UNet performed best in standard image quality measures, such as MAE, the contour based anatomical feature comparison with the CBCT showed that Dual-UNet performed worst on anatomical comparison. This emphasizes the importance of adding anatomy based evaluation of sCTs generated by dl-models. For applications in the pelvic area, direct anatomical comparison with the CBCT may provide a useful method to assess the clinical applicability of dl-based sCT generation methods. |
first_indexed | 2024-04-10T00:29:30Z |
format | Article |
id | doaj.art-12e7ece56588480a9289459c8d1938fe |
institution | Directory Open Access Journal |
issn | 2405-6316 |
language | English |
last_indexed | 2024-04-10T00:29:30Z |
publishDate | 2023-01-01 |
publisher | Elsevier |
record_format | Article |
series | Physics and Imaging in Radiation Oncology |
spelling | doaj.art-12e7ece56588480a9289459c8d1938fe2023-03-15T04:28:40ZengElsevierPhysics and Imaging in Radiation Oncology2405-63162023-01-0125100416Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomographyYvonne J.M. de Hond0Camiel E.M. Kerckhaert1Maureen A.J.M. van Eijnatten2Paul M.A. van Haaren3Coen W. Hurkmans4Rob H.N. Tijssen5Department of Radiation Oncology, Catharina Hospital, Eindhoven, The Netherlands; Corresponding author.Department of Radiation Oncology, Catharina Hospital, Eindhoven, The Netherlands; Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, The NetherlandsDepartment of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, The NetherlandsDepartment of Radiation Oncology, Catharina Hospital, Eindhoven, The NetherlandsDepartment of Radiation Oncology, Catharina Hospital, Eindhoven, The NetherlandsDepartment of Radiation Oncology, Catharina Hospital, Eindhoven, The NetherlandsBackground and purpose: To improve cone-beam computed tomography (CBCT), deep-learning (DL)-models are being explored to generate synthetic CTs (sCT). The sCT evaluation is mainly focused on image quality and CT number accuracy. However, correct representation of daily anatomy of the CBCT is also important for sCTs in adaptive radiotherapy. The aim of this study was to emphasize the importance of anatomical correctness by quantitatively assessing sCT scans generated from CBCT scans using different paired and unpaired dl-models. Materials and methods: Planning CTs (pCT) and CBCTs of 56 prostate cancer patients were included to generate sCTs. Three different dl-models, Dual-UNet, Single-UNet and Cycle-consistent Generative Adversarial Network (CycleGAN), were evaluated on image quality and anatomical correctness. The image quality was assessed using image metrics, such as Mean Absolute Error (MAE). The anatomical correctness between sCT and CBCT was quantified using organs-at-risk volumes and average surface distances (ASD). Results: MAE was 24 Hounsfield Unit (HU) [range:19-30 HU] for Dual-UNet, 40 HU [range:34-56 HU] for Single-UNet and 41HU [range:37-46 HU] for CycleGAN. Bladder ASD was 4.5 mm [range:1.6–12.3 mm] for Dual-UNet, 0.7 mm [range:0.4–1.2 mm] for Single-UNet and 0.9 mm [range:0.4–1.1 mm] CycleGAN. Conclusions: Although Dual-UNet performed best in standard image quality measures, such as MAE, the contour based anatomical feature comparison with the CBCT showed that Dual-UNet performed worst on anatomical comparison. This emphasizes the importance of adding anatomy based evaluation of sCTs generated by dl-models. For applications in the pelvic area, direct anatomical comparison with the CBCT may provide a useful method to assess the clinical applicability of dl-based sCT generation methods.http://www.sciencedirect.com/science/article/pii/S2405631623000076Deep-LearningSynthetic computed tomographyCone-beam computed tomographyAverage surface distanceAnatomical comparisonVolume difference |
spellingShingle | Yvonne J.M. de Hond Camiel E.M. Kerckhaert Maureen A.J.M. van Eijnatten Paul M.A. van Haaren Coen W. Hurkmans Rob H.N. Tijssen Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography Physics and Imaging in Radiation Oncology Deep-Learning Synthetic computed tomography Cone-beam computed tomography Average surface distance Anatomical comparison Volume difference |
title | Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography |
title_full | Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography |
title_fullStr | Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography |
title_full_unstemmed | Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography |
title_short | Anatomical evaluation of deep-learning synthetic computed tomography images generated from male pelvis cone-beam computed tomography |
title_sort | anatomical evaluation of deep learning synthetic computed tomography images generated from male pelvis cone beam computed tomography |
topic | Deep-Learning Synthetic computed tomography Cone-beam computed tomography Average surface distance Anatomical comparison Volume difference |
url | http://www.sciencedirect.com/science/article/pii/S2405631623000076 |
work_keys_str_mv | AT yvonnejmdehond anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography AT camielemkerckhaert anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography AT maureenajmvaneijnatten anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography AT paulmavanhaaren anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography AT coenwhurkmans anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography AT robhntijssen anatomicalevaluationofdeeplearningsyntheticcomputedtomographyimagesgeneratedfrommalepelvisconebeamcomputedtomography |