Impact of pixel-based machine-learning techniques on automated frameworks for delineation of gross tumor volume regions for stereotactic body radiation therapy

Yasuo Kawata, Hidetaka Arimura, Koujirou Ikushima, Ze Jin, Kento Morita, Chiaki Tokunaga, Hidetake Yabuuchi, Yoshiyuki Shioyama, Tomonari Sasaki, Hiroshi Honda, Masayuki Sasaki

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

The aim of this study was to investigate the impact of pixel-based machine learning (ML) techniques, i.e., fuzzy-c-means clustering method (FCM), and the artificial neural network (ANN) and support vector machine (SVM), on an automated framework for delineation of gross tumor volume (GTV) regions of lung cancer for stereotactic body radiation therapy. The morphological and metabolic features for GTV regions, which were determined based on the knowledge of radiation oncologists, were fed on a pixel-by-pixel basis into the respective FCM, ANN, and SVM ML techniques. Then, the ML techniques were incorporated into the automated delineation framework of GTVs followed by an optimum contour selection (OCS) method, which we proposed in a previous study. The three-ML-based frameworks were evaluated for 16 lung cancer cases (six solid, four ground glass opacity (GGO), six part-solid GGO) with the datasets of planning computed tomography (CT) and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/CT images using the three-dimensional Dice similarity coefficient (DSC). DSC denotes the degree of region similarity between the GTVs contoured by radiation oncologists and those estimated using the automated framework. The FCM-based framework achieved the highest DSCs of 0.79 ± 0.06, whereas DSCs of the ANN-based and SVM-based frameworks were 0.76 ± 0.14 and 0.73 ± 0.14, respectively. The FCM-based framework provided the highest segmentation accuracy and precision without a learning process (lowest calculation cost). Therefore, the FCM-based framework can be useful for delineation of tumor regions in practical treatment planning.

Original languageEnglish
Pages (from-to)141-149
Number of pages9
JournalPhysica Medica
Volume42
DOIs
Publication statusPublished - Oct 1 2017

Fingerprint

Body Regions
machine learning
delineation
Tumor Burden
radiation therapy
Radiotherapy
tumors
pixels
Cluster Analysis
tomography
opacity
lungs
planning
cancer
Glass
Lung Neoplasms
glass
radiation
coefficients
Three-Dimensional Imaging

All Science Journal Classification (ASJC) codes

  • Biophysics
  • Radiology Nuclear Medicine and imaging
  • Physics and Astronomy(all)

Cite this

Impact of pixel-based machine-learning techniques on automated frameworks for delineation of gross tumor volume regions for stereotactic body radiation therapy. / Kawata, Yasuo; Arimura, Hidetaka; Ikushima, Koujirou; Jin, Ze; Morita, Kento; Tokunaga, Chiaki; Yabuuchi, Hidetake; Shioyama, Yoshiyuki; Sasaki, Tomonari; Honda, Hiroshi; Sasaki, Masayuki.

In: Physica Medica, Vol. 42, 01.10.2017, p. 141-149.

Research output: Contribution to journalArticle

@article{77386827105a493eb2ba36ff9cd039c4,
title = "Impact of pixel-based machine-learning techniques on automated frameworks for delineation of gross tumor volume regions for stereotactic body radiation therapy",
abstract = "The aim of this study was to investigate the impact of pixel-based machine learning (ML) techniques, i.e., fuzzy-c-means clustering method (FCM), and the artificial neural network (ANN) and support vector machine (SVM), on an automated framework for delineation of gross tumor volume (GTV) regions of lung cancer for stereotactic body radiation therapy. The morphological and metabolic features for GTV regions, which were determined based on the knowledge of radiation oncologists, were fed on a pixel-by-pixel basis into the respective FCM, ANN, and SVM ML techniques. Then, the ML techniques were incorporated into the automated delineation framework of GTVs followed by an optimum contour selection (OCS) method, which we proposed in a previous study. The three-ML-based frameworks were evaluated for 16 lung cancer cases (six solid, four ground glass opacity (GGO), six part-solid GGO) with the datasets of planning computed tomography (CT) and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/CT images using the three-dimensional Dice similarity coefficient (DSC). DSC denotes the degree of region similarity between the GTVs contoured by radiation oncologists and those estimated using the automated framework. The FCM-based framework achieved the highest DSCs of 0.79 ± 0.06, whereas DSCs of the ANN-based and SVM-based frameworks were 0.76 ± 0.14 and 0.73 ± 0.14, respectively. The FCM-based framework provided the highest segmentation accuracy and precision without a learning process (lowest calculation cost). Therefore, the FCM-based framework can be useful for delineation of tumor regions in practical treatment planning.",
author = "Yasuo Kawata and Hidetaka Arimura and Koujirou Ikushima and Ze Jin and Kento Morita and Chiaki Tokunaga and Hidetake Yabuuchi and Yoshiyuki Shioyama and Tomonari Sasaki and Hiroshi Honda and Masayuki Sasaki",
year = "2017",
month = "10",
day = "1",
doi = "10.1016/j.ejmp.2017.08.012",
language = "English",
volume = "42",
pages = "141--149",
journal = "Physica Medica",
issn = "1120-1797",
publisher = "Associazione Italiana di Fisica Medica",

}

TY - JOUR

T1 - Impact of pixel-based machine-learning techniques on automated frameworks for delineation of gross tumor volume regions for stereotactic body radiation therapy

AU - Kawata, Yasuo

AU - Arimura, Hidetaka

AU - Ikushima, Koujirou

AU - Jin, Ze

AU - Morita, Kento

AU - Tokunaga, Chiaki

AU - Yabuuchi, Hidetake

AU - Shioyama, Yoshiyuki

AU - Sasaki, Tomonari

AU - Honda, Hiroshi

AU - Sasaki, Masayuki

PY - 2017/10/1

Y1 - 2017/10/1

N2 - The aim of this study was to investigate the impact of pixel-based machine learning (ML) techniques, i.e., fuzzy-c-means clustering method (FCM), and the artificial neural network (ANN) and support vector machine (SVM), on an automated framework for delineation of gross tumor volume (GTV) regions of lung cancer for stereotactic body radiation therapy. The morphological and metabolic features for GTV regions, which were determined based on the knowledge of radiation oncologists, were fed on a pixel-by-pixel basis into the respective FCM, ANN, and SVM ML techniques. Then, the ML techniques were incorporated into the automated delineation framework of GTVs followed by an optimum contour selection (OCS) method, which we proposed in a previous study. The three-ML-based frameworks were evaluated for 16 lung cancer cases (six solid, four ground glass opacity (GGO), six part-solid GGO) with the datasets of planning computed tomography (CT) and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/CT images using the three-dimensional Dice similarity coefficient (DSC). DSC denotes the degree of region similarity between the GTVs contoured by radiation oncologists and those estimated using the automated framework. The FCM-based framework achieved the highest DSCs of 0.79 ± 0.06, whereas DSCs of the ANN-based and SVM-based frameworks were 0.76 ± 0.14 and 0.73 ± 0.14, respectively. The FCM-based framework provided the highest segmentation accuracy and precision without a learning process (lowest calculation cost). Therefore, the FCM-based framework can be useful for delineation of tumor regions in practical treatment planning.

AB - The aim of this study was to investigate the impact of pixel-based machine learning (ML) techniques, i.e., fuzzy-c-means clustering method (FCM), and the artificial neural network (ANN) and support vector machine (SVM), on an automated framework for delineation of gross tumor volume (GTV) regions of lung cancer for stereotactic body radiation therapy. The morphological and metabolic features for GTV regions, which were determined based on the knowledge of radiation oncologists, were fed on a pixel-by-pixel basis into the respective FCM, ANN, and SVM ML techniques. Then, the ML techniques were incorporated into the automated delineation framework of GTVs followed by an optimum contour selection (OCS) method, which we proposed in a previous study. The three-ML-based frameworks were evaluated for 16 lung cancer cases (six solid, four ground glass opacity (GGO), six part-solid GGO) with the datasets of planning computed tomography (CT) and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/CT images using the three-dimensional Dice similarity coefficient (DSC). DSC denotes the degree of region similarity between the GTVs contoured by radiation oncologists and those estimated using the automated framework. The FCM-based framework achieved the highest DSCs of 0.79 ± 0.06, whereas DSCs of the ANN-based and SVM-based frameworks were 0.76 ± 0.14 and 0.73 ± 0.14, respectively. The FCM-based framework provided the highest segmentation accuracy and precision without a learning process (lowest calculation cost). Therefore, the FCM-based framework can be useful for delineation of tumor regions in practical treatment planning.

UR - http://www.scopus.com/inward/record.url?scp=85035088805&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85035088805&partnerID=8YFLogxK

U2 - 10.1016/j.ejmp.2017.08.012

DO - 10.1016/j.ejmp.2017.08.012

M3 - Article

C2 - 29173908

AN - SCOPUS:85035088805

VL - 42

SP - 141

EP - 149

JO - Physica Medica

JF - Physica Medica

SN - 1120-1797

ER -