Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area
This study investigated the feasibility of deep learning-based segmentation (DLS) and continual training for adaptive radiotherapy (RT) of head and neck (H&N) cancer. One-hundred patients treated with definitive RT were included. Based on 23 organs-at-risk (OARs) manually segmented in initial pl...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2021-02-01
|
Series: | Cancers |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-6694/13/4/702 |
id |
doaj-946afc8cba0d4cacb8c981163d36255d |
---|---|
record_format |
Article |
spelling |
doaj-946afc8cba0d4cacb8c981163d36255d2021-02-10T00:04:18ZengMDPI AGCancers2072-66942021-02-011370270210.3390/cancers13040702Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck AreaNalee Kim0Jaehee Chun1Jee Suk Chang2Chang Geol Lee3Ki Chang Keum4Jin Sung Kim5Department of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaDepartment of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaDepartment of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaDepartment of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaDepartment of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaDepartment of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul 03722, KoreaThis study investigated the feasibility of deep learning-based segmentation (DLS) and continual training for adaptive radiotherapy (RT) of head and neck (H&N) cancer. One-hundred patients treated with definitive RT were included. Based on 23 organs-at-risk (OARs) manually segmented in initial planning computed tomography (CT), modified FC-DenseNet was trained for DLS: (i) using data obtained from 60 patients, with 20 matched patients in the test set (DLSm); (ii) using data obtained from 60 identical patients with 20 unmatched patients in the test set (DLSu). Manually contoured OARs in adaptive planning CT for independent 20 patients were provided as test sets. Deformable image registration (DIR) was also performed. All 23 OARs were compared using quantitative measurements, and nine OARs were also evaluated via subjective assessment from 26 observers using the Turing test. DLSm achieved better performance than both DLSu and DIR (mean Dice similarity coefficient; 0.83 vs. 0.80 vs. 0.70), mainly for glandular structures, whose volume significantly reduced during RT. Based on subjective measurements, DLS is often perceived as a human (49.2%). Furthermore, DLSm is preferred over DLSu (67.2%) and DIR (96.7%), with a similar rate of required revision to that of manual segmentation (28.0% vs. 29.7%). In conclusion, DLS was effective and preferred over DIR. Additionally, continual DLS training is required for an effective optimization and robustness in personalized adaptive RT.https://www.mdpi.com/2072-6694/13/4/702head and neck cancerdeep learningauto segmentationartificial intelligenceadaptive radiation therapy |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Nalee Kim Jaehee Chun Jee Suk Chang Chang Geol Lee Ki Chang Keum Jin Sung Kim |
spellingShingle |
Nalee Kim Jaehee Chun Jee Suk Chang Chang Geol Lee Ki Chang Keum Jin Sung Kim Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area Cancers head and neck cancer deep learning auto segmentation artificial intelligence adaptive radiation therapy |
author_facet |
Nalee Kim Jaehee Chun Jee Suk Chang Chang Geol Lee Ki Chang Keum Jin Sung Kim |
author_sort |
Nalee Kim |
title |
Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area |
title_short |
Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area |
title_full |
Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area |
title_fullStr |
Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area |
title_full_unstemmed |
Feasibility of Continual Deep Learning-Based Segmentation for Personalized Adaptive Radiation Therapy in Head and Neck Area |
title_sort |
feasibility of continual deep learning-based segmentation for personalized adaptive radiation therapy in head and neck area |
publisher |
MDPI AG |
series |
Cancers |
issn |
2072-6694 |
publishDate |
2021-02-01 |
description |
This study investigated the feasibility of deep learning-based segmentation (DLS) and continual training for adaptive radiotherapy (RT) of head and neck (H&N) cancer. One-hundred patients treated with definitive RT were included. Based on 23 organs-at-risk (OARs) manually segmented in initial planning computed tomography (CT), modified FC-DenseNet was trained for DLS: (i) using data obtained from 60 patients, with 20 matched patients in the test set (DLSm); (ii) using data obtained from 60 identical patients with 20 unmatched patients in the test set (DLSu). Manually contoured OARs in adaptive planning CT for independent 20 patients were provided as test sets. Deformable image registration (DIR) was also performed. All 23 OARs were compared using quantitative measurements, and nine OARs were also evaluated via subjective assessment from 26 observers using the Turing test. DLSm achieved better performance than both DLSu and DIR (mean Dice similarity coefficient; 0.83 vs. 0.80 vs. 0.70), mainly for glandular structures, whose volume significantly reduced during RT. Based on subjective measurements, DLS is often perceived as a human (49.2%). Furthermore, DLSm is preferred over DLSu (67.2%) and DIR (96.7%), with a similar rate of required revision to that of manual segmentation (28.0% vs. 29.7%). In conclusion, DLS was effective and preferred over DIR. Additionally, continual DLS training is required for an effective optimization and robustness in personalized adaptive RT. |
topic |
head and neck cancer deep learning auto segmentation artificial intelligence adaptive radiation therapy |
url |
https://www.mdpi.com/2072-6694/13/4/702 |
work_keys_str_mv |
AT naleekim feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea AT jaeheechun feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea AT jeesukchang feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea AT changgeollee feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea AT kichangkeum feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea AT jinsungkim feasibilityofcontinualdeeplearningbasedsegmentationforpersonalizedadaptiveradiationtherapyinheadandneckarea |
_version_ |
1724275811698081792 |