Deep-learning based surface region selection for deep inspiration breath hold (DIBH) monitoring in left breast cancer radiotherapy

Haibin Chen, Mingli Chen, Weiguo Lu, Bo Zhao, Steve B Jiang, Linghong Zhou, Nathan Kim, Ann E Spangler, Assal S Rahimi, Xin Zhen, Xuejun Gu

Research output: Contribution to journalArticle

Abstract

Deep inspiration breath hold (DIBH) with surface supervising is a common technique for cardiac dose reduction in left breast cancer radiotherapy. Surface supervision accuracy relies on the characteristics of surface region. In this study, a convolutional neural network (CNN) based automatic region-of-interest (ROI) selection method was proposed to select an optimal surface ROI for DIBH surface monitoring. The curvature entropy and the normal of each vertex on the breast cancer patient surface were calculated and formed as representative maps for ROI selection learning. 900 ROIs were randomly extracted from each patient's surface representative map, and the corresponding rigid ROI registration errors (REs) were calculated. The VGG-16 (a 16-layer network structure developed by Visual Geometry Group(VGG) from University of Oxford) pre-trained on a large natural image database ImageNet were fine-tuned using 27 thousand extracted ROIs and the corresponding REs from thirty patients. The RE prediction accuracy of the trained model was validated on additional ten patients. Satisfactory RE predictive accuracies were achieved with the root mean square error (RMSE)/mean absolute error (MAE) smaller than 1 mm/0.7 mm in translations and 0.45/0.35 in rotations, respectively. The REs of the model selected ROIs on ten testing cases is close to the minimal predicted RE with mean RE differences <1 mm and <0.5 for translation and rotation, respectively. The proposed RE predictive model can be utilized for selecting a quasi-optimal ROI in left breast cancer DIBH radiotherapy (DIBH-RT).

Original languageEnglish (US)
Article number245013
JournalPhysics in Medicine and Biology
Volume63
Issue number24
DOIs
StatePublished - Dec 12 2018

Fingerprint

Radiotherapy
Learning
Breast Neoplasms
Patient Advocacy
Entropy
Databases

Keywords

  • DIBH
  • motion monitoring
  • ROI selection
  • transfer learning

ASJC Scopus subject areas

  • Radiological and Ultrasound Technology
  • Radiology Nuclear Medicine and imaging

Cite this

Deep-learning based surface region selection for deep inspiration breath hold (DIBH) monitoring in left breast cancer radiotherapy. / Chen, Haibin; Chen, Mingli; Lu, Weiguo; Zhao, Bo; Jiang, Steve B; Zhou, Linghong; Kim, Nathan; Spangler, Ann E; Rahimi, Assal S; Zhen, Xin; Gu, Xuejun.

In: Physics in Medicine and Biology, Vol. 63, No. 24, 245013, 12.12.2018.

Research output: Contribution to journalArticle

@article{a0845c14e2e6422eb7c59d989a6b9f85,
title = "Deep-learning based surface region selection for deep inspiration breath hold (DIBH) monitoring in left breast cancer radiotherapy",
abstract = "Deep inspiration breath hold (DIBH) with surface supervising is a common technique for cardiac dose reduction in left breast cancer radiotherapy. Surface supervision accuracy relies on the characteristics of surface region. In this study, a convolutional neural network (CNN) based automatic region-of-interest (ROI) selection method was proposed to select an optimal surface ROI for DIBH surface monitoring. The curvature entropy and the normal of each vertex on the breast cancer patient surface were calculated and formed as representative maps for ROI selection learning. 900 ROIs were randomly extracted from each patient's surface representative map, and the corresponding rigid ROI registration errors (REs) were calculated. The VGG-16 (a 16-layer network structure developed by Visual Geometry Group(VGG) from University of Oxford) pre-trained on a large natural image database ImageNet were fine-tuned using 27 thousand extracted ROIs and the corresponding REs from thirty patients. The RE prediction accuracy of the trained model was validated on additional ten patients. Satisfactory RE predictive accuracies were achieved with the root mean square error (RMSE)/mean absolute error (MAE) smaller than 1 mm/0.7 mm in translations and 0.45/0.35 in rotations, respectively. The REs of the model selected ROIs on ten testing cases is close to the minimal predicted RE with mean RE differences <1 mm and <0.5 for translation and rotation, respectively. The proposed RE predictive model can be utilized for selecting a quasi-optimal ROI in left breast cancer DIBH radiotherapy (DIBH-RT).",
keywords = "DIBH, motion monitoring, ROI selection, transfer learning",
author = "Haibin Chen and Mingli Chen and Weiguo Lu and Bo Zhao and Jiang, {Steve B} and Linghong Zhou and Nathan Kim and Spangler, {Ann E} and Rahimi, {Assal S} and Xin Zhen and Xuejun Gu",
year = "2018",
month = "12",
day = "12",
doi = "10.1088/1361-6560/aaf0d6",
language = "English (US)",
volume = "63",
journal = "Physics in Medicine and Biology",
issn = "0031-9155",
publisher = "IOP Publishing Ltd.",
number = "24",

}

TY - JOUR

T1 - Deep-learning based surface region selection for deep inspiration breath hold (DIBH) monitoring in left breast cancer radiotherapy

AU - Chen, Haibin

AU - Chen, Mingli

AU - Lu, Weiguo

AU - Zhao, Bo

AU - Jiang, Steve B

AU - Zhou, Linghong

AU - Kim, Nathan

AU - Spangler, Ann E

AU - Rahimi, Assal S

AU - Zhen, Xin

AU - Gu, Xuejun

PY - 2018/12/12

Y1 - 2018/12/12

N2 - Deep inspiration breath hold (DIBH) with surface supervising is a common technique for cardiac dose reduction in left breast cancer radiotherapy. Surface supervision accuracy relies on the characteristics of surface region. In this study, a convolutional neural network (CNN) based automatic region-of-interest (ROI) selection method was proposed to select an optimal surface ROI for DIBH surface monitoring. The curvature entropy and the normal of each vertex on the breast cancer patient surface were calculated and formed as representative maps for ROI selection learning. 900 ROIs were randomly extracted from each patient's surface representative map, and the corresponding rigid ROI registration errors (REs) were calculated. The VGG-16 (a 16-layer network structure developed by Visual Geometry Group(VGG) from University of Oxford) pre-trained on a large natural image database ImageNet were fine-tuned using 27 thousand extracted ROIs and the corresponding REs from thirty patients. The RE prediction accuracy of the trained model was validated on additional ten patients. Satisfactory RE predictive accuracies were achieved with the root mean square error (RMSE)/mean absolute error (MAE) smaller than 1 mm/0.7 mm in translations and 0.45/0.35 in rotations, respectively. The REs of the model selected ROIs on ten testing cases is close to the minimal predicted RE with mean RE differences <1 mm and <0.5 for translation and rotation, respectively. The proposed RE predictive model can be utilized for selecting a quasi-optimal ROI in left breast cancer DIBH radiotherapy (DIBH-RT).

AB - Deep inspiration breath hold (DIBH) with surface supervising is a common technique for cardiac dose reduction in left breast cancer radiotherapy. Surface supervision accuracy relies on the characteristics of surface region. In this study, a convolutional neural network (CNN) based automatic region-of-interest (ROI) selection method was proposed to select an optimal surface ROI for DIBH surface monitoring. The curvature entropy and the normal of each vertex on the breast cancer patient surface were calculated and formed as representative maps for ROI selection learning. 900 ROIs were randomly extracted from each patient's surface representative map, and the corresponding rigid ROI registration errors (REs) were calculated. The VGG-16 (a 16-layer network structure developed by Visual Geometry Group(VGG) from University of Oxford) pre-trained on a large natural image database ImageNet were fine-tuned using 27 thousand extracted ROIs and the corresponding REs from thirty patients. The RE prediction accuracy of the trained model was validated on additional ten patients. Satisfactory RE predictive accuracies were achieved with the root mean square error (RMSE)/mean absolute error (MAE) smaller than 1 mm/0.7 mm in translations and 0.45/0.35 in rotations, respectively. The REs of the model selected ROIs on ten testing cases is close to the minimal predicted RE with mean RE differences <1 mm and <0.5 for translation and rotation, respectively. The proposed RE predictive model can be utilized for selecting a quasi-optimal ROI in left breast cancer DIBH radiotherapy (DIBH-RT).

KW - DIBH

KW - motion monitoring

KW - ROI selection

KW - transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85058745771&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058745771&partnerID=8YFLogxK

U2 - 10.1088/1361-6560/aaf0d6

DO - 10.1088/1361-6560/aaf0d6

M3 - Article

C2 - 30523967

AN - SCOPUS:85058745771

VL - 63

JO - Physics in Medicine and Biology

JF - Physics in Medicine and Biology

SN - 0031-9155

IS - 24

M1 - 245013

ER -