Use of machine learning methods for prediction of acute toxicity in organs at risk following prostate radiotherapy

Research output: Contribution to journalArticle

Abstract

Purpose: The goal of this study is to investigate the advantages of large scale optimization methods vs conventional classification techniques in predicting acute toxicity for urinary bladder and rectum due to prostate irradiation. Methods: Clinical and dosimetric data of 321 patients undergoing prostate conformal radiotherapy were recorded. Gastro-intestinal and genito-urinary acute toxicities were scored according to the Radiation Therapy Oncology GroupEuropean Organization for Research and Treatment of Cancer (RTOGEORTC) scale. Patients were classified in two categories to separate mild (Grade2) from severe toxicity levels (Grade2). Machine learning methods at different complexity were implemented to predict toxicity as a function of multiple variables. The first approach consisted of a large scale optimization method, based on genetic algorithms (GAs) and artificial neural networks (ANN). The second approach was a binary classifier based on support vector machines (SVM). Results: The ANN and SVM-based solutions showed comparable prediction accuracy, exhibiting an area under the receiver operating characteristic (ROC) curve of 0.7. Different sensitivity and specificity features were measured for the two approaches. The ANN algorithm showed enhanced sensitivity if combined with appropriate classification criteria. Conclusions: The results demonstrate that high sensitivity in toxicity prediction can be achieved with optimized ANNs, that are put forward to represent a valuable support in medical decisions. Future studies will be focused on enlarging the available patient database to increase the reliability of toxicity prediction algorithms and to define optimal classification criteria.

Original languageEnglish
Pages (from-to)2859-2867
Number of pages9
JournalMedical Physics
Volume38
Issue number6
DOIs
Publication statusPublished - Jun 2011

Fingerprint

Organs at Risk
Prostate
Radiotherapy
Conformal Radiotherapy
Radiation Oncology
Rectum
ROC Curve
Urinary Bladder
Databases
Sensitivity and Specificity
Machine Learning
Research
Neoplasms
Support Vector Machine

Keywords

  • artificial neural networks
  • machine learning in medicine
  • prostate cancer
  • radiotherapy
  • support vector machine

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

@article{20b37167239d4f858b25595b60ea80f7,
title = "Use of machine learning methods for prediction of acute toxicity in organs at risk following prostate radiotherapy",
abstract = "Purpose: The goal of this study is to investigate the advantages of large scale optimization methods vs conventional classification techniques in predicting acute toxicity for urinary bladder and rectum due to prostate irradiation. Methods: Clinical and dosimetric data of 321 patients undergoing prostate conformal radiotherapy were recorded. Gastro-intestinal and genito-urinary acute toxicities were scored according to the Radiation Therapy Oncology GroupEuropean Organization for Research and Treatment of Cancer (RTOGEORTC) scale. Patients were classified in two categories to separate mild (Grade2) from severe toxicity levels (Grade2). Machine learning methods at different complexity were implemented to predict toxicity as a function of multiple variables. The first approach consisted of a large scale optimization method, based on genetic algorithms (GAs) and artificial neural networks (ANN). The second approach was a binary classifier based on support vector machines (SVM). Results: The ANN and SVM-based solutions showed comparable prediction accuracy, exhibiting an area under the receiver operating characteristic (ROC) curve of 0.7. Different sensitivity and specificity features were measured for the two approaches. The ANN algorithm showed enhanced sensitivity if combined with appropriate classification criteria. Conclusions: The results demonstrate that high sensitivity in toxicity prediction can be achieved with optimized ANNs, that are put forward to represent a valuable support in medical decisions. Future studies will be focused on enlarging the available patient database to increase the reliability of toxicity prediction algorithms and to define optimal classification criteria.",
keywords = "artificial neural networks, machine learning in medicine, prostate cancer, radiotherapy, support vector machine",
author = "Andrea Pella and Raffaella Cambria and Marco Riboldi and Jereczek-Fossa, {Barbara Alicja} and Cristiana Fodor and Dario Zerini and Torshabi, {Ahmad Esmaili} and Federica Cattani and Cristina Garibaldi and Guido Pedroli and Guido Baroni and Roberto Orecchia",
year = "2011",
month = "6",
doi = "10.1118/1.3582947",
language = "English",
volume = "38",
pages = "2859--2867",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "John Wiley and Sons Ltd",
number = "6",

}

TY - JOUR

T1 - Use of machine learning methods for prediction of acute toxicity in organs at risk following prostate radiotherapy

AU - Pella, Andrea

AU - Cambria, Raffaella

AU - Riboldi, Marco

AU - Jereczek-Fossa, Barbara Alicja

AU - Fodor, Cristiana

AU - Zerini, Dario

AU - Torshabi, Ahmad Esmaili

AU - Cattani, Federica

AU - Garibaldi, Cristina

AU - Pedroli, Guido

AU - Baroni, Guido

AU - Orecchia, Roberto

PY - 2011/6

Y1 - 2011/6

N2 - Purpose: The goal of this study is to investigate the advantages of large scale optimization methods vs conventional classification techniques in predicting acute toxicity for urinary bladder and rectum due to prostate irradiation. Methods: Clinical and dosimetric data of 321 patients undergoing prostate conformal radiotherapy were recorded. Gastro-intestinal and genito-urinary acute toxicities were scored according to the Radiation Therapy Oncology GroupEuropean Organization for Research and Treatment of Cancer (RTOGEORTC) scale. Patients were classified in two categories to separate mild (Grade2) from severe toxicity levels (Grade2). Machine learning methods at different complexity were implemented to predict toxicity as a function of multiple variables. The first approach consisted of a large scale optimization method, based on genetic algorithms (GAs) and artificial neural networks (ANN). The second approach was a binary classifier based on support vector machines (SVM). Results: The ANN and SVM-based solutions showed comparable prediction accuracy, exhibiting an area under the receiver operating characteristic (ROC) curve of 0.7. Different sensitivity and specificity features were measured for the two approaches. The ANN algorithm showed enhanced sensitivity if combined with appropriate classification criteria. Conclusions: The results demonstrate that high sensitivity in toxicity prediction can be achieved with optimized ANNs, that are put forward to represent a valuable support in medical decisions. Future studies will be focused on enlarging the available patient database to increase the reliability of toxicity prediction algorithms and to define optimal classification criteria.

AB - Purpose: The goal of this study is to investigate the advantages of large scale optimization methods vs conventional classification techniques in predicting acute toxicity for urinary bladder and rectum due to prostate irradiation. Methods: Clinical and dosimetric data of 321 patients undergoing prostate conformal radiotherapy were recorded. Gastro-intestinal and genito-urinary acute toxicities were scored according to the Radiation Therapy Oncology GroupEuropean Organization for Research and Treatment of Cancer (RTOGEORTC) scale. Patients were classified in two categories to separate mild (Grade2) from severe toxicity levels (Grade2). Machine learning methods at different complexity were implemented to predict toxicity as a function of multiple variables. The first approach consisted of a large scale optimization method, based on genetic algorithms (GAs) and artificial neural networks (ANN). The second approach was a binary classifier based on support vector machines (SVM). Results: The ANN and SVM-based solutions showed comparable prediction accuracy, exhibiting an area under the receiver operating characteristic (ROC) curve of 0.7. Different sensitivity and specificity features were measured for the two approaches. The ANN algorithm showed enhanced sensitivity if combined with appropriate classification criteria. Conclusions: The results demonstrate that high sensitivity in toxicity prediction can be achieved with optimized ANNs, that are put forward to represent a valuable support in medical decisions. Future studies will be focused on enlarging the available patient database to increase the reliability of toxicity prediction algorithms and to define optimal classification criteria.

KW - artificial neural networks

KW - machine learning in medicine

KW - prostate cancer

KW - radiotherapy

KW - support vector machine

UR - http://www.scopus.com/inward/record.url?scp=79958798015&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79958798015&partnerID=8YFLogxK

U2 - 10.1118/1.3582947

DO - 10.1118/1.3582947

M3 - Article

C2 - 21815361

AN - SCOPUS:79958798015

VL - 38

SP - 2859

EP - 2867

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 6

ER -