Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients

Image-guided treatment adaptation is a game changer in oncological particle therapy (PT), especially for younger patients. The purpose of this study is to present a cycle generative adversarial network (CycleGAN)-based method for synthetic computed tomography (sCT) generation from cone beam CT (CBCT...

Full description

Saved in:
Bibliographic Details
Main Authors: Matteo Pepa, Siavash Taleghani, Giulia Sellaro, Alfredo Mirandola, Francesca Colombo, Sabina Vennarini, Mario Ciocca, Chiara Paganelli, Ester Orlandi, Guido Baroni, Andrea Pella
Format: Article
Language:English
Published: MDPI AG 2024-11-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/24/23/7460
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846123859582910464
author Matteo Pepa
Siavash Taleghani
Giulia Sellaro
Alfredo Mirandola
Francesca Colombo
Sabina Vennarini
Mario Ciocca
Chiara Paganelli
Ester Orlandi
Guido Baroni
Andrea Pella
author_facet Matteo Pepa
Siavash Taleghani
Giulia Sellaro
Alfredo Mirandola
Francesca Colombo
Sabina Vennarini
Mario Ciocca
Chiara Paganelli
Ester Orlandi
Guido Baroni
Andrea Pella
author_sort Matteo Pepa
collection DOAJ
description Image-guided treatment adaptation is a game changer in oncological particle therapy (PT), especially for younger patients. The purpose of this study is to present a cycle generative adversarial network (CycleGAN)-based method for synthetic computed tomography (sCT) generation from cone beam CT (CBCT) towards adaptive PT (APT) of paediatric patients. Firstly, 44 CBCTs of 15 young pelvic patients were pre-processed to reduce ring artefacts and rigidly registered on same-day CT scans (i.e., verification CT scans, vCT scans) and then inputted to the CycleGAN network (employing either Res-Net and U-Net generators) to synthesise sCT. In particular, 36 and 8 volumes were used for training and testing, respectively. Image quality was evaluated qualitatively and quantitatively using the structural similarity index metric (SSIM) and the peak signal-to-noise ratio (PSNR) between registered CBCT (rCBCT) and vCT and between sCT and vCT to evaluate the improvements brought by CycleGAN. Despite limitations due to the sub-optimal input image quality and the small field of view (FOV), the quality of sCT was found to be overall satisfactory from a quantitative and qualitative perspective. Our findings indicate that CycleGAN is promising to produce sCT scans with acceptable CT-like image texture in paediatric settings, even when CBCT with narrow fields of view (FOV) are employed.
format Article
id doaj-art-bc947236dbf04a5e987a4205c1bb7293
institution Kabale University
issn 1424-8220
language English
publishDate 2024-11-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-bc947236dbf04a5e987a4205c1bb72932024-12-13T16:31:36ZengMDPI AGSensors1424-82202024-11-012423746010.3390/s24237460Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric PatientsMatteo Pepa0Siavash Taleghani1Giulia Sellaro2Alfredo Mirandola3Francesca Colombo4Sabina Vennarini5Mario Ciocca6Chiara Paganelli7Ester Orlandi8Guido Baroni9Andrea Pella10Bioengineering Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyDepartment of Electronics, Information and Bioengineering, Politecnico di Milano (POLIMI), 20133 Milan, ItalyBioengineering Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyMedical Physics Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyRadiation Oncology Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyPaediatric Radiotherapy Unit, Fondazione IRCCS Istituto Nazionale dei Tumori (INT), 20133 Milan, ItalyMedical Physics Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyDepartment of Electronics, Information and Bioengineering, Politecnico di Milano (POLIMI), 20133 Milan, ItalyRadiation Oncology Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyDepartment of Electronics, Information and Bioengineering, Politecnico di Milano (POLIMI), 20133 Milan, ItalyBioengineering Unit, Clinical Department, CNAO National Centre for Oncological Hadrontherapy, 27100 Pavia, ItalyImage-guided treatment adaptation is a game changer in oncological particle therapy (PT), especially for younger patients. The purpose of this study is to present a cycle generative adversarial network (CycleGAN)-based method for synthetic computed tomography (sCT) generation from cone beam CT (CBCT) towards adaptive PT (APT) of paediatric patients. Firstly, 44 CBCTs of 15 young pelvic patients were pre-processed to reduce ring artefacts and rigidly registered on same-day CT scans (i.e., verification CT scans, vCT scans) and then inputted to the CycleGAN network (employing either Res-Net and U-Net generators) to synthesise sCT. In particular, 36 and 8 volumes were used for training and testing, respectively. Image quality was evaluated qualitatively and quantitatively using the structural similarity index metric (SSIM) and the peak signal-to-noise ratio (PSNR) between registered CBCT (rCBCT) and vCT and between sCT and vCT to evaluate the improvements brought by CycleGAN. Despite limitations due to the sub-optimal input image quality and the small field of view (FOV), the quality of sCT was found to be overall satisfactory from a quantitative and qualitative perspective. Our findings indicate that CycleGAN is promising to produce sCT scans with acceptable CT-like image texture in paediatric settings, even when CBCT with narrow fields of view (FOV) are employed.https://www.mdpi.com/1424-8220/24/23/7460proton therapycarbon ion therapyadaptive particle therapypaediatric oncologyCBCTsynthetic CT
spellingShingle Matteo Pepa
Siavash Taleghani
Giulia Sellaro
Alfredo Mirandola
Francesca Colombo
Sabina Vennarini
Mario Ciocca
Chiara Paganelli
Ester Orlandi
Guido Baroni
Andrea Pella
Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
Sensors
proton therapy
carbon ion therapy
adaptive particle therapy
paediatric oncology
CBCT
synthetic CT
title Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
title_full Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
title_fullStr Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
title_full_unstemmed Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
title_short Unsupervised Deep Learning for Synthetic CT Generation from CBCT Images for Proton and Carbon Ion Therapy for Paediatric Patients
title_sort unsupervised deep learning for synthetic ct generation from cbct images for proton and carbon ion therapy for paediatric patients
topic proton therapy
carbon ion therapy
adaptive particle therapy
paediatric oncology
CBCT
synthetic CT
url https://www.mdpi.com/1424-8220/24/23/7460
work_keys_str_mv AT matteopepa unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT siavashtaleghani unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT giuliasellaro unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT alfredomirandola unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT francescacolombo unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT sabinavennarini unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT mariociocca unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT chiarapaganelli unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT esterorlandi unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT guidobaroni unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients
AT andreapella unsuperviseddeeplearningforsyntheticctgenerationfromcbctimagesforprotonandcarboniontherapyforpaediatricpatients