Prediction of IDH1 Mutation Status in Glioblastoma Using Machine Learning Technique Based on Quantitative Radiomic Data

doi:10.1016/j.wneu.2019.01.157

World Neurosurgery

Volume 125, May 2019, Pages e688-e696

https://doi.org/10.1016/j.wneu.2019.01.157 Get rights and content

Objective

Isocitrate dehydrogenase 1 (IDH1) mutation status is an independent favorable prognostic factor for glioblastoma (GBM) and is usually determined by sequencing or immunohistochemistry. An accurate prediction of IDH1 mutation status via noninvasive methods helps establish the appropriate treatment strategy. We aimed to predict IDH1 mutation status using quantitative radiomic data in patients with GBM.

Methods

Between May 2010 and June 2015, we retrospectively identified 88 patients with newly diagnosed GBM. After semiautomatic segmentation of the lesions, we extracted 31 features from preoperative multiparametric magnetic resonance images. We also determined IDH1 mutation status using targeted sequencing and immunohistochemistry. A training cohort (n = 88) was used to train machine learning−based classifiers, with internal validation. The machine-learning technique was then validated in an external dataset of 35 patients with GBM.

Results

We detected the IDH1 mutation in 12 out of 88 GBMs. Multiparametric radiomic profiles revealed that the IDH1 mutation was associated with a smaller enhancing area volume and a larger necrotic area volume. Using the machine learning−based classification algorithms, we identified 70.3%−87.3% of prediction rate of IDH1 mutation status and found 66.3%−83.4% accuracy in the external validation set.

Conclusions

We demonstrate that machine learning algorithms can predict IDH1 mutation status in GBM using preoperative multiparametric magnetic resonance images.

Introduction

Quantitative radiomic data can be used to extract additional information from multiparametric magnetic resonance images. The field of radiogenomics has recently been established to study the relationship between radiologic image features and molecular characteristics.1, 2, 3, 4 Radiologic imaging techniques provide opportunities to study unique data for different types of tissue. Several explorative studies have demonstrated associations between molecular characteristics and imaging features.2, 3, 5, 6, 7, 8, 9, 10, 11, 12 Quantitative radiomic data can provide new information from multiparametric magnetic resonance imaging (MRI). New strategies for imaging-based computer-aided diagnostics and therapies have been widely explored. In recent years, machine-learning techniques have become major tools for medical image analysis in various computer-aided diagnostic applications.¹³ We have previously reported a comprehensive radiogenomic study of glioblastoma (GBM), in which quantitative large-scale radiomic profiling was integrated with genomic data.¹⁴

Mutation of the isocitrate dehydrogenase 1 (IDH1) gene occurs in up to 12% of GBMs.15, 16 Tumors with IDH1 mutations (IDHmut) are associated with improved outcomes when compared with those with IDH1 wildtype (IDHwt).17, 18, 19 IDHmut GBM cases may be either primary or secondary GBMs.²⁰ Therefore it is important to determine IDH1 mutation status, which can now be achieved using sequencing or immunohistochemistry techniques. To reduce the cost of testing and the risk of surgery, noninvasive methods for the accurate prediction of IDH1 mutation status have been widely investigated. The aim of this study was to use machine learning−based classification models to predict IDH1 mutation status based on preoperative MRI features.

Section snippets

Patient Enrolment

Between May 2010 and June 2015, a cohort of 88 cases was selected from the data registry of the Department of Neurosurgery, Samsung Medical Center. All patients met the following criteria: 1) previously untreated and histologically confirmed grade IV GBM, according to the World Health Organization classification; 2) available clinical variables including patient demographics; 3) available Sanger sequencing or immunohistochemistry results for IDH1 mutation detection; and 4) available

Patient Characteristics

Between May 2010 and June 2015, we retrospectively identified 88 patients with treatment-naïve GBM and available clinical, pathologic, and radiologic information. The median age of the patients at the time of operation was 52 years (range, 20–80 years). The population comprised 47 men and 41 women. The median duration of clinical follow-up after the operation was 52.6 weeks (range, 3.7–240.0 weeks). IDH1 mutations were identified in 12 of the 88 cases. The mutation was confirmed by

Discussion

To better understand the pathophysiology of cancer and to improve treatment outcomes, there has been increasing emphasis on the identification of molecular phenotypes. According to the 2016 World Health Organization Classification of Tumors of the Central Nervous System,26, 27 GBMs are now divided into IDH-wild type and IDH-mutant GBMs. IDH1 gene mutations are associated with longer survival of patients with GBM relative to wild-type IDH1.15, 16, 28, 29, 30 Therefore early identification of IDH1

References (48)

W.B. Pope
Genomics of brain tumor imaging
Neuroimaging Clin N Am
(2015)
P. Lambin et al.
Radiomics: extracting more information from medical images using advanced feature analysis
Eur J Cancer
(2012)
O. Gevaert et al.
Glioblastoma multiforme: exploratory radiogenomic analysis by using quantitative image features
Radiology
(2014)
B.M. Ellingson
Radiogenomics and imaging phenotypes in glioblastoma: novel observations and correlation with molecular characteristics
Curr Neurol Neurosci Rep
(2015)
P.O. Zinn et al.
Radiogenomic mapping of edema/cellular invasion MRI-phenotypes in glioblastoma multiforme
PLoS One
(2011)
B.M. Ellingson et al.
Probabilistic radiographic atlas of glioblastoma phenotypes
AJNR Am J Neuroradiol
(2013)
K.M. Naeini et al.
Identifying the mesenchymal molecular subtype of glioblastoma using quantitative volumetric analysis of anatomic magnetic resonance images
Neuro Oncol
(2013)
D.A. Gutman et al.
MR imaging predictors of molecular profile and survival: multi-institutional study of the TCGA glioblastoma data set
Radiology
(2013)
M. Diehn et al.
Identification of noninvasive imaging surrogates for brain tumor gene-expression modules
Proc Natl Acad Sci U S A
(2008)
W.B. Pope et al.
Relationship between gene expression and enhancement in glioblastoma multiforme: exploratory DNA microarray analysis
Radiology
(2008)

R.R. Colen et al.

Imaging genomic mapping of an invasive MRI phenotype predicts patient outcome and metabolic dysfunction: a TCGA glioma phenotype research group project

BMC Med Genomics

(2014)

M.R. Carlson et al.

Relationship between survival and edema in malignant gliomas: role of vascular endothelial growth factor and neuronal pentraxin 2

Clin Cancer Res

(2007)

W.B. Pope et al.

Differential gene expression in glioblastoma defined by ADC histogram analysis: relationship to extracellular matrix molecules and survival

AJNR Am J Neuroradiol

(2012)

M. Mohri et al.

Foundations of Machine Learning

(2012)

D.S. Kong et al.

Integrative radiogenomic analysis for multicentric radiophenotype in glioblastoma

Oncotarget

(2016)

H. Yan et al.

IDH1 and IDH2 mutations in gliomas

N Engl J Med

(2009)

D.W. Parsons et al.

An integrated genomic analysis of human glioblastoma multiforme

Science

(2008)

S. Nobusawa et al.

IDH1 mutations as molecular signature and predictive factor of secondary glioblastomas

Clin Cancer Res

(2009)

C. Hartmann et al.

Patients with IDH1 wild type anaplastic astrocytomas exhibit worse prognosis than IDH1-mutated glioblastomas, and IDH1 mutation status accounts for the unfavorable prognostic effect of higher age: implications for classification of gliomas

Acta Neuropathol

(2010)

M. Sanson et al.

Isocitrate dehydrogenase 1 codon 132 mutation is an important prognostic biomarker in gliomas

J Clin Oncol

(2009)

K. Ichimura et al.

IDH1 mutations are present in the majority of common adult gliomas but rare in primary glioblastomas

Neuro Oncol

(2009)

S.C. Jung et al.

Cerebral blood volume analysis in glioblastomas using dynamic susceptibility contrast-enhanced perfusion MRI: a comparison of manual and semiautomatic segmentation methods

PLoS One

(2013)

B.M. Ellingson et al.

Pretreatment ADC histogram analysis is a predictive imaging biomarker for bevacizumab treatment but not chemotherapy in recurrent glioblastoma

AJNR Am J Neuroradiol

(2014)

W.B. Pope et al.

Recurrent glioblastoma multiforme: ADC histogram analysis predicts response to bevacizumab treatment

Radiology

(2009)

Cited by (32)

A hierarchical machine learning model based on Glioblastoma patients' clinical, biomedical, and image data to analyze their treatment plans
2022, Computers in Biology and Medicine
Citation Excerpt :
Their model's final accuracy and F1-score with optimal threshold were 87% and 74%, respectively. [31] applied a new machine learning method to predict Isocitrate dehydrogenase-I mutation status based on Radiomic data of GBM patients. They studied 31 extracted features from MRI of 88 patients with GBM in the neurosurgery department, Samsung medical center, between 2010 and 2015.
Glioblastoma Multiforme (GBM) is an aggressive brain cancer in adults that kills most patients in the first year due to ineffective treatment. Different clinical, biomedical, and image data features are needed to analyze GBM, increasing complexities. Besides, they lead to weak performances for machine learning models due to ignoring physicians' knowledge. Therefore, this paper proposes a hierarchical model based on Fuzzy C-mean (FCM) clustering, Wrapper feature selection, and twelve classifiers to analyze treatment plans.
The proposed method finds the effectiveness of previous and current treatment plans, hierarchically determining the best decision for future treatment plans for GBM patients using clinical data, biomedical data, and different image data. A case study is presented based on the Cancer Genome Atlas Glioblastoma Multiforme dataset to prove the effectiveness of the proposed model. This dataset is analyzed using data preprocessing, experts' knowledge, and a feature reduction method based on the Principal Component Analysis. Then, the FCM clustering method is utilized to reinforce classifier learning.
The proposed model finds the best combination of Wrapper feature selection and classifier for each cluster based on different measures, including accuracy, sensitivity, specificity, precision, F-score, and G-mean according to a hierarchical structure. It has the best performance among other reinforced classifiers. Besides, this model is compatible with real-world medical processes for GBM patients based on clinical, biomedical, and image data.
A Review of Radiomics and Deep Predictive Modeling in Glioma Characterization
2021, Academic Radiology
Recent developments in glioma categorization based on biological genotypes and application of computational machine learning or deep learning based predictive models using multi-modal MRI biomarkers to assess these genotypes provides potential assurance for optimal and personalized treatment plans and efficacy. Artificial intelligence based quantified assessment of glioma using MRI derived hand-crafted or auto-extracted features have become crucial as genomic alterations can be associated with MRI based phenotypes. This survey integrates all the recent work carried out in state-of-the-art radiomics, and Artificial Intelligence based learning solutions related to molecular diagnosis, prognosis, and treatment monitoring with the aim to create a structured resource on radiogenomic analysis of glioma. Challenges such as inter-scanner variability, requirement of benchmark datasets, prospective validations for clinical applicability are discussed with further scope for designing optimal solutions for glioma stratification with immediate recommendations for further diagnostic decisions and personalized treatment plans for glioma patients.
Machine learning applications to neuroimaging for glioma detection and classification: An artificial intelligence augmented systematic review
2021, Journal of Clinical Neuroscience
Glioma is the most common primary intraparenchymal tumor of the brain and the 5-year survival rate of high-grade glioma is poor. Magnetic resonance imaging (MRI) is essential for detecting, characterizing and monitoring brain tumors but definitive diagnosis still relies on surgical pathology. Machine learning has been applied to the analysis of MRI data in glioma research and has the potential to change clinical practice and improve patient outcomes. This systematic review synthesizes and analyzes the current state of machine learning applications to glioma MRI data and explores the use of machine learning for systematic review automation. Various datapoints were extracted from the 153 studies that met inclusion criteria and analyzed. Natural language processing (NLP) analysis involved keyword extraction, topic modeling and document classification. Machine learning has been applied to tumor grading and diagnosis, tumor segmentation, non-invasive genomic biomarker identification, detection of progression and patient survival prediction. Model performance was generally strong (AUC = 0.87 ± 0.09; sensitivity = 0.87 ± 0.10; specificity = 0.0.86 ± 0.10; precision = 0.88 ± 0.11). Convolutional neural network, support vector machine and random forest algorithms were top performers. Deep learning document classifiers yielded acceptable performance (mean 5-fold cross-validation AUC = 0.71). Machine learning tools and data resources were synthesized and summarized to facilitate future research. Machine learning has been widely applied to the processing of MRI data in glioma research and has demonstrated substantial utility. NLP and transfer learning resources enabled the successful development of a replicable method for automating the systematic review article screening process, which has potential for shortening the time from discovery to clinical application in medicine.
A radiomics–clinical nomogram for preoperative prediction of IDH1 mutation in primary glioblastoma multiforme
2020, Clinical Radiology
Citation Excerpt :
The benefit to patient management at the early stage of the disease and in follow-up of the non-invasive prediction of the IDH mutation status of glioma using MRI has been demonstrated in several previous studies.9,10,14 A recent study reported the prediction of the IDH1 mutation status in GBM based on 31 features from preoperative multiparametric MRI images,12 but they did not take radiomics features into consideration nor did they incorporate clinical features into the model. Another radiomics study on the IDH1 mutation in low-grade glioma also showed that a stratifying strategy helped to predict the IDH1 mutation status; unfortunately, however, the relatively small patient population (only 57 patients) and the absence of a validation cohort reduced its power.13
To develop and validate an individualised radiomics–clinical nomogram for the prediction of the isocitrate dehydrogenase 1 (IDH1) mutation status in primary glioblastoma multiforme (GBM) based on radiomics features and clinical variables.
In a retrospective study, preoperative magnetic resonance imaging (MRI) images were obtained of 122 patients with primary glioblastoma (development cohort = 101; validation cohort = 21). Radiomics features were extracted from total tumour based on the post-contrast high-resolution three-dimensional (3D) T1-weighted MRI images. Radiomics features were selected by using a least absolute shrinkage and selection operator (LASSO) binomial regression model with nested cross-validation. Then, a radiomics–clinical nomogram was constructed by combining relevant radiomics features and clinical variables and subsequently tested by using the independent validation cohort.
A total of 105 features were quantified on the 3D MRI images of each patient, and eight were selected to construct the radiomics model for predicting IDH1 mutation status. The mean classification accuracy and mean κ value achieved with the model were 88.4±3% and 0.701±0.08, respectively. The radiomics–clinical nomogram, which combines eight radiomics features and three clinical variables (patient age, sex and tumour location), demonstrated good discrimination (C-index 0.934 [95% CI, 0.874 to 0.994]; F1 score 0.78) and performed well with the validation cohort (C-index 0.963 [95% CI, 0.957 to 0.969]; F1 score 0.91; AUC 0.956).
A radiomics–clinical nomogram was developed and proved to be valuable in the non-invasive, individualised prediction of the IDH1 mutation status in patients with primary GBM. The nomogram can be applied using clinical conditions to facilitate preoperative patient evaluation.
IDH1 mutation prediction using MR-based radiomics in glioblastoma: comparison between manual and fully automated deep learning-based approach of tumor segmentation
2020, European Journal of Radiology
Citation Excerpt :
At the same time, T2WI was reported to reflect heterogeneity of glioblastoma and was capable of differentiating it from pseudo-progression [26]. The results of current study are also consistent with another similar study by Lee et al. [27], who implemented radiomic analysis to predict IDH1 mutation status in glioblastoma. While our IDH1 predictive performance was similar to their study (Lee et al.: 70.3–87.3 % prediction rate vs. current study: 75.8–86.8 %), they used only manual segmentation of multi-parametric MRI.
This study aimed to determine whether MR-based radiomics of glioblastoma can predict the isocitrate dehydrogenase-1 (IDH1) mutation status and compare predictive performances between manual and fully automatic deep-learning segmentations.
Forty-five glioblastoma patients with pretreatment T2-weighted MRIs were retrospectively evaluated. Manual segmentations of glioblastoma and peri-tumoral edema were trained via a deep neural network (V-Net). An independent external cohort of 137 glioblastoma patients from the Cancer Imaging Archive was also included (test set 1, n = 46; test set 2, n = 91). Test set 1—without known IDH1 status—was used to calculate dice similarity coefficients (DSC) between the two segmentation methods (manual & V-Net). From test set 2, all-relevant radiomic features were selected via a random forest-based wrapper algorithm for IDH1 prediction. Receiver operating characteristics (ROC) curves with areas under the curve (AUC) were plotted as performance metrics for both methods.
Among 136 patients (45 and 91 patients from our institution and test set 2, respectively), 17 patients (11.2 %) had IDH1 mutations. The mean DSC of test set 1 was 0.78 ± 0.14 (range, 0.34−0.94). A subset of 9 all-relevant features (8.4 %, 9/107) was selected. V-Net segmentation of the test set 2 yielded similar performance in predicting IDH1 mutation as compared to manual segmentation (V-Net AUC = 0.86 vs. manual AUC = 0.90). The optimal cut-point threshold of AUC yielded 86.8 % accuracy for manual segmentation and 75.8 % for V-Net segmentation.
V-Net showed robust segmentation capability of glioblastoma on T2-weighted MRI. All-relevant radiomics features from both segmentation methods yielded a similar performance in IDH1 prediction.
Artificial intelligence-based MRI radiomics and radiogenomics in glioma
2024, Cancer Imaging

View all citing articles on Scopus

: Conflict of interest statement: This research was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (HI14C3418), and by a grant (NRF-2015M3A9A7029740) from the National Research Foundation, funded by the Ministry of Science, ICT, and Future Planning (MSIP) of Korea. There are no known conflicts of interest associated with this publication.

View full text

Original ArticlePrediction of IDH1 Mutation Status in Glioblastoma Using Machine Learning Technique Based on Quantitative Radiomic Data

Objective

Methods

Results

Conclusions

Introduction

Section snippets

Patient Enrolment

Patient Characteristics

Discussion

Neuroimaging Clin N Am

Eur J Cancer

Glioblastoma multiforme: exploratory radiogenomic analysis by using quantitative image features

Radiology

Radiogenomics and imaging phenotypes in glioblastoma: novel observations and correlation with molecular characteristics

Curr Neurol Neurosci Rep

Radiogenomic mapping of edema/cellular invasion MRI-phenotypes in glioblastoma multiforme

PLoS One

Probabilistic radiographic atlas of glioblastoma phenotypes

AJNR Am J Neuroradiol

Identifying the mesenchymal molecular subtype of glioblastoma using quantitative volumetric analysis of anatomic magnetic resonance images

Neuro Oncol

MR imaging predictors of molecular profile and survival: multi-institutional study of the TCGA glioblastoma data set

Radiology

Identification of noninvasive imaging surrogates for brain tumor gene-expression modules

Proc Natl Acad Sci U S A

Relationship between gene expression and enhancement in glioblastoma multiforme: exploratory DNA microarray analysis

Radiology

Imaging genomic mapping of an invasive MRI phenotype predicts patient outcome and metabolic dysfunction: a TCGA glioma phenotype research group project

BMC Med Genomics

Relationship between survival and edema in malignant gliomas: role of vascular endothelial growth factor and neuronal pentraxin 2

Clin Cancer Res

Differential gene expression in glioblastoma defined by ADC histogram analysis: relationship to extracellular matrix molecules and survival

AJNR Am J Neuroradiol

Foundations of Machine Learning

Integrative radiogenomic analysis for multicentric radiophenotype in glioblastoma

Oncotarget

IDH1 and IDH2 mutations in gliomas

N Engl J Med

An integrated genomic analysis of human glioblastoma multiforme

Science

IDH1 mutations as molecular signature and predictive factor of secondary glioblastomas

Clin Cancer Res

Patients with IDH1 wild type anaplastic astrocytomas exhibit worse prognosis than IDH1-mutated glioblastomas, and IDH1 mutation status accounts for the unfavorable prognostic effect of higher age: implications for classification of gliomas

Acta Neuropathol

Isocitrate dehydrogenase 1 codon 132 mutation is an important prognostic biomarker in gliomas

J Clin Oncol

IDH1 mutations are present in the majority of common adult gliomas but rare in primary glioblastomas

Neuro Oncol

Cerebral blood volume analysis in glioblastomas using dynamic susceptibility contrast-enhanced perfusion MRI: a comparison of manual and semiautomatic segmentation methods

PLoS One

Pretreatment ADC histogram analysis is a predictive imaging biomarker for bevacizumab treatment but not chemotherapy in recurrent glioblastoma

AJNR Am J Neuroradiol

Recurrent glioblastoma multiforme: ADC histogram analysis predicts response to bevacizumab treatment

Radiology

Original Article
Prediction of IDH1 Mutation Status in Glioblastoma Using Machine Learning Technique Based on Quantitative Radiomic Data