Multi-label Nonlinear Matrix Completion with Transductive Multi-task Feature Selection for Joint MGMT and IDH1 Status Prediction of Patient with High-Grade Gliomas

Lei Chen, Han Zhang, Junfeng Lu, Kimhan Thung, Abudumijiti Aibaidula, Luyan Liu, Songcan Chen, Lei Jin, Jinsong Wu, Qian Wang, Liangfu Zhou, Dinggang Shen

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

The O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation and isocitrate dehydrogenase 1 (IDH1) mutation in high-grade gliomas (HGG) have proven to be the two important molecular indicators associated with better prognosis. Traditionally, the statuses of MGMT and IDH1 are obtained via surgical biopsy, which has limited their wider clinical implementation. Accurate presurgical prediction of their statuses based on preoperative multimodal neuroimaging is of great clinical value for a better treatment plan. Currently, the available dataset associated with this study has several challenges, such as small sample size and complex, nonlinear (image) feature-to-(molecular) label relationship. To address these issues, we propose a novel Multi-label Nonlinear Matrix Completion (MNMC) model to jointly predict both MGMT and IDH1 statuses in a multi-task framework. Specifically, we first employ a nonlinear random Fourier feature mapping to improve the linear separability of the data, and then use transductive multi-task feature selection (performed in a nonlinearly transformed feature space) to refine the imputed soft labels, thus alleviating the overfitting problem caused by small sample size. We further design an optimization algorithm with a guaranteed convergence ability based on a block prox-linear method to solve the proposed MNMC model. Finally, by using a single-center, multimodal brain imaging and molecular pathology dataset of HGG, we derive brain functional and structural connectomics features to jointly predict MGMT and IDH1 statuses. Results demonstrate that our proposed method outperforms the previously widely used single- and multi-task machine learning methods. This study also shows the promise of utilizing brain connectomics for HGG prognosis in a non-invasive manner.

Original languageEnglish
JournalIEEE Transactions on Medical Imaging
DOIs
Publication statusAccepted/In press - 2018 Feb 16
Externally publishedYes

Fingerprint

Isocitrate Dehydrogenase
Methyltransferases
Glioma
Feature extraction
Labels
Connectome
DNA
Joints
Brain
Neuroimaging
Sample Size
Multimodal Imaging
Naproxen
Methylation
Molecular Pathology
Biopsy
Pathology
Learning systems
Imaging techniques
Mutation

Keywords

  • Brain
  • Brain tumor
  • connectomics
  • Electronic mail
  • Feature extraction
  • functional connectivity
  • high-grade glioma
  • Imaging
  • matrix completion
  • molecular biomarker
  • Pathology
  • prognosis
  • structural connectivity
  • Testing
  • Tumors

ASJC Scopus subject areas

  • Software
  • Radiological and Ultrasound Technology
  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this

Multi-label Nonlinear Matrix Completion with Transductive Multi-task Feature Selection for Joint MGMT and IDH1 Status Prediction of Patient with High-Grade Gliomas. / Chen, Lei; Zhang, Han; Lu, Junfeng; Thung, Kimhan; Aibaidula, Abudumijiti; Liu, Luyan; Chen, Songcan; Jin, Lei; Wu, Jinsong; Wang, Qian; Zhou, Liangfu; Shen, Dinggang.

In: IEEE Transactions on Medical Imaging, 16.02.2018.

Research output: Contribution to journalArticle

Chen, Lei ; Zhang, Han ; Lu, Junfeng ; Thung, Kimhan ; Aibaidula, Abudumijiti ; Liu, Luyan ; Chen, Songcan ; Jin, Lei ; Wu, Jinsong ; Wang, Qian ; Zhou, Liangfu ; Shen, Dinggang. / Multi-label Nonlinear Matrix Completion with Transductive Multi-task Feature Selection for Joint MGMT and IDH1 Status Prediction of Patient with High-Grade Gliomas. In: IEEE Transactions on Medical Imaging. 2018.
@article{6e8046f26ea949c7b9c8bc15db05cf7a,
title = "Multi-label Nonlinear Matrix Completion with Transductive Multi-task Feature Selection for Joint MGMT and IDH1 Status Prediction of Patient with High-Grade Gliomas",
abstract = "The O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation and isocitrate dehydrogenase 1 (IDH1) mutation in high-grade gliomas (HGG) have proven to be the two important molecular indicators associated with better prognosis. Traditionally, the statuses of MGMT and IDH1 are obtained via surgical biopsy, which has limited their wider clinical implementation. Accurate presurgical prediction of their statuses based on preoperative multimodal neuroimaging is of great clinical value for a better treatment plan. Currently, the available dataset associated with this study has several challenges, such as small sample size and complex, nonlinear (image) feature-to-(molecular) label relationship. To address these issues, we propose a novel Multi-label Nonlinear Matrix Completion (MNMC) model to jointly predict both MGMT and IDH1 statuses in a multi-task framework. Specifically, we first employ a nonlinear random Fourier feature mapping to improve the linear separability of the data, and then use transductive multi-task feature selection (performed in a nonlinearly transformed feature space) to refine the imputed soft labels, thus alleviating the overfitting problem caused by small sample size. We further design an optimization algorithm with a guaranteed convergence ability based on a block prox-linear method to solve the proposed MNMC model. Finally, by using a single-center, multimodal brain imaging and molecular pathology dataset of HGG, we derive brain functional and structural connectomics features to jointly predict MGMT and IDH1 statuses. Results demonstrate that our proposed method outperforms the previously widely used single- and multi-task machine learning methods. This study also shows the promise of utilizing brain connectomics for HGG prognosis in a non-invasive manner.",
keywords = "Brain, Brain tumor, connectomics, Electronic mail, Feature extraction, functional connectivity, high-grade glioma, Imaging, matrix completion, molecular biomarker, Pathology, prognosis, structural connectivity, Testing, Tumors",
author = "Lei Chen and Han Zhang and Junfeng Lu and Kimhan Thung and Abudumijiti Aibaidula and Luyan Liu and Songcan Chen and Lei Jin and Jinsong Wu and Qian Wang and Liangfu Zhou and Dinggang Shen",
year = "2018",
month = "2",
day = "16",
doi = "10.1109/TMI.2018.2807590",
language = "English",
journal = "IEEE Transactions on Medical Imaging",
issn = "0278-0062",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Multi-label Nonlinear Matrix Completion with Transductive Multi-task Feature Selection for Joint MGMT and IDH1 Status Prediction of Patient with High-Grade Gliomas

AU - Chen, Lei

AU - Zhang, Han

AU - Lu, Junfeng

AU - Thung, Kimhan

AU - Aibaidula, Abudumijiti

AU - Liu, Luyan

AU - Chen, Songcan

AU - Jin, Lei

AU - Wu, Jinsong

AU - Wang, Qian

AU - Zhou, Liangfu

AU - Shen, Dinggang

PY - 2018/2/16

Y1 - 2018/2/16

N2 - The O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation and isocitrate dehydrogenase 1 (IDH1) mutation in high-grade gliomas (HGG) have proven to be the two important molecular indicators associated with better prognosis. Traditionally, the statuses of MGMT and IDH1 are obtained via surgical biopsy, which has limited their wider clinical implementation. Accurate presurgical prediction of their statuses based on preoperative multimodal neuroimaging is of great clinical value for a better treatment plan. Currently, the available dataset associated with this study has several challenges, such as small sample size and complex, nonlinear (image) feature-to-(molecular) label relationship. To address these issues, we propose a novel Multi-label Nonlinear Matrix Completion (MNMC) model to jointly predict both MGMT and IDH1 statuses in a multi-task framework. Specifically, we first employ a nonlinear random Fourier feature mapping to improve the linear separability of the data, and then use transductive multi-task feature selection (performed in a nonlinearly transformed feature space) to refine the imputed soft labels, thus alleviating the overfitting problem caused by small sample size. We further design an optimization algorithm with a guaranteed convergence ability based on a block prox-linear method to solve the proposed MNMC model. Finally, by using a single-center, multimodal brain imaging and molecular pathology dataset of HGG, we derive brain functional and structural connectomics features to jointly predict MGMT and IDH1 statuses. Results demonstrate that our proposed method outperforms the previously widely used single- and multi-task machine learning methods. This study also shows the promise of utilizing brain connectomics for HGG prognosis in a non-invasive manner.

AB - The O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation and isocitrate dehydrogenase 1 (IDH1) mutation in high-grade gliomas (HGG) have proven to be the two important molecular indicators associated with better prognosis. Traditionally, the statuses of MGMT and IDH1 are obtained via surgical biopsy, which has limited their wider clinical implementation. Accurate presurgical prediction of their statuses based on preoperative multimodal neuroimaging is of great clinical value for a better treatment plan. Currently, the available dataset associated with this study has several challenges, such as small sample size and complex, nonlinear (image) feature-to-(molecular) label relationship. To address these issues, we propose a novel Multi-label Nonlinear Matrix Completion (MNMC) model to jointly predict both MGMT and IDH1 statuses in a multi-task framework. Specifically, we first employ a nonlinear random Fourier feature mapping to improve the linear separability of the data, and then use transductive multi-task feature selection (performed in a nonlinearly transformed feature space) to refine the imputed soft labels, thus alleviating the overfitting problem caused by small sample size. We further design an optimization algorithm with a guaranteed convergence ability based on a block prox-linear method to solve the proposed MNMC model. Finally, by using a single-center, multimodal brain imaging and molecular pathology dataset of HGG, we derive brain functional and structural connectomics features to jointly predict MGMT and IDH1 statuses. Results demonstrate that our proposed method outperforms the previously widely used single- and multi-task machine learning methods. This study also shows the promise of utilizing brain connectomics for HGG prognosis in a non-invasive manner.

KW - Brain

KW - Brain tumor

KW - connectomics

KW - Electronic mail

KW - Feature extraction

KW - functional connectivity

KW - high-grade glioma

KW - Imaging

KW - matrix completion

KW - molecular biomarker

KW - Pathology

KW - prognosis

KW - structural connectivity

KW - Testing

KW - Tumors

UR - http://www.scopus.com/inward/record.url?scp=85042190438&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85042190438&partnerID=8YFLogxK

U2 - 10.1109/TMI.2018.2807590

DO - 10.1109/TMI.2018.2807590

M3 - Article

C2 - 29994582

AN - SCOPUS:85042190438

JO - IEEE Transactions on Medical Imaging

JF - IEEE Transactions on Medical Imaging

SN - 0278-0062

ER -