Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status

Research output: Contribution to journalArticle

26 Citations (Scopus)

Abstract

Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/− 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/− 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/− 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.

Original languageEnglish (US)
Pages (from-to)1-7
Number of pages7
JournalJournal of Digital Imaging
DOIs
StateAccepted/In press - Aug 7 2017

Fingerprint

Methylation
Methyltransferases
Tumors
Neural networks
Biomarkers
Network architecture
Brain Neoplasms
Magnetic resonance imaging
Brain
Neoplasms
Genes
Imaging techniques

Keywords

  • Deep learning
  • MGMT methylation
  • MRI

ASJC Scopus subject areas

  • Radiological and Ultrasound Technology
  • Radiology Nuclear Medicine and imaging
  • Computer Science Applications

Cite this

@article{61ae34f0590345bf88897f2b3e4aa4a7,
title = "Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status",
abstract = "Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90{\%} (+/− 3.92{\%}) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72{\%} (+/− 13.61{\%}) while ResNet18 (18 layers) accuracy was 76.75{\%} (+/− 20.67{\%}). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.",
keywords = "Deep learning, MGMT methylation, MRI",
author = "Panagiotis Korfiatis and Timothy Kline and Lachance, {Daniel H} and Parney, {Ian F} and Buckner, {Jan Craig} and Erickson, {Bradley J}",
year = "2017",
month = "8",
day = "7",
doi = "10.1007/s10278-017-0009-z",
language = "English (US)",
pages = "1--7",
journal = "Journal of Digital Imaging",
issn = "0897-1889",
publisher = "Springer New York",

}

TY - JOUR

T1 - Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status

AU - Korfiatis, Panagiotis

AU - Kline, Timothy

AU - Lachance, Daniel H

AU - Parney, Ian F

AU - Buckner, Jan Craig

AU - Erickson, Bradley J

PY - 2017/8/7

Y1 - 2017/8/7

N2 - Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/− 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/− 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/− 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.

AB - Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/− 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/− 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/− 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.

KW - Deep learning

KW - MGMT methylation

KW - MRI

UR - http://www.scopus.com/inward/record.url?scp=85026904407&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85026904407&partnerID=8YFLogxK

U2 - 10.1007/s10278-017-0009-z

DO - 10.1007/s10278-017-0009-z

M3 - Article

SP - 1

EP - 7

JO - Journal of Digital Imaging

JF - Journal of Digital Imaging

SN - 0897-1889

ER -