Robust training of recurrent neural networks to handle missing data for disease progression modeling

Publikation: KonferencebidragPaperForskning

Standard

Robust training of recurrent neural networks to handle missing data for disease progression modeling. / Mehdipour Ghazi, Mostafa; Nielsen, Mads; Pai, Akshay Sadananda Uppinakudru; Cardoso, Manuel Jorge; Modat, Marc; Ourselin, Sebastien; Sørensen, Lauge.

2018. Paper præsenteret ved 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Holland.

Publikation: KonferencebidragPaperForskning

Harvard

Mehdipour Ghazi, M, Nielsen, M, Pai, ASU, Cardoso, MJ, Modat, M, Ourselin, S & Sørensen, L 2018, 'Robust training of recurrent neural networks to handle missing data for disease progression modeling', Paper fremlagt ved 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Holland, 04/07/2018 - 06/07/2018. <https://openreview.net/forum?id=S1jqMb2oM>

APA

Mehdipour Ghazi, M., Nielsen, M., Pai, A. S. U., Cardoso, M. J., Modat, M., Ourselin, S., & Sørensen, L. (2018). Robust training of recurrent neural networks to handle missing data for disease progression modeling. Paper præsenteret ved 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Holland. https://openreview.net/forum?id=S1jqMb2oM

Vancouver

Mehdipour Ghazi M, Nielsen M, Pai ASU, Cardoso MJ, Modat M, Ourselin S o.a.. Robust training of recurrent neural networks to handle missing data for disease progression modeling. 2018. Paper præsenteret ved 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Holland.

Author

Mehdipour Ghazi, Mostafa ; Nielsen, Mads ; Pai, Akshay Sadananda Uppinakudru ; Cardoso, Manuel Jorge ; Modat, Marc ; Ourselin, Sebastien ; Sørensen, Lauge. / Robust training of recurrent neural networks to handle missing data for disease progression modeling. Paper præsenteret ved 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Holland.9 s.

Bibtex

@conference{b04c7b5a415c4bb2801a6f2355c93bee,
title = "Robust training of recurrent neural networks to handle missing data for disease progression modeling",
abstract = "Disease progression modeling (DPM) using longitudinal data is a challenging task in machine learning for healthcare that can provide clinicians with better tools for diagnosis and monitoring of disease. Existing DPM algorithms neglect temporal dependencies among measurements and make parametric assumptions about biomarker trajectories. In addition, they do not model multiple biomarkers jointly and need to align subjects' trajectories. In this paper, recurrent neural networks (RNNs) are utilized to address these issues. However, in many cases, longitudinal cohorts contain incomplete data, which hinders the application of standard RNNs and requires a pre-processing step such as imputation of the missing values. We, therefore, propose a generalized training rule for the most widely used RNN architecture, long short-term memory (LSTM) networks, that can handle missing values in both target and predictor variables. This algorithm is applied for modeling the progression of Alzheimer's disease (AD) using magnetic resonance imaging (MRI) biomarkers. The results show that the proposed LSTM algorithm achieves a lower mean absolute error for prediction of measurements across all considered MRI biomarkers compared to using standard LSTM networks with data imputation or using a regression-based DPM method. Moreover, applying linear discriminant analysis to the biomarkers' values predicted by the proposed algorithm results in a larger area under the receiver operating characteristic curve (AUC) for clinical diagnosis of AD compared to the same alternatives, and the AUC is comparable to state-of-the-art AUC's from a recent cross-sectional medical image classification challenge. This paper shows that built-in handling of missing values in LSTM network training paves the way for application of RNNs in disease progression modeling.",
author = "{Mehdipour Ghazi}, Mostafa and Mads Nielsen and Pai, {Akshay Sadananda Uppinakudru} and Cardoso, {Manuel Jorge} and Marc Modat and Sebastien Ourselin and Lauge S{\o}rensen",
year = "2018",
language = "English",
note = "1st Conference on Medical Imaging with Deep Learning (MIDL 2018) ; Conference date: 04-07-2018 Through 06-07-2018",

}

RIS

TY - CONF

T1 - Robust training of recurrent neural networks to handle missing data for disease progression modeling

AU - Mehdipour Ghazi, Mostafa

AU - Nielsen, Mads

AU - Pai, Akshay Sadananda Uppinakudru

AU - Cardoso, Manuel Jorge

AU - Modat, Marc

AU - Ourselin, Sebastien

AU - Sørensen, Lauge

PY - 2018

Y1 - 2018

N2 - Disease progression modeling (DPM) using longitudinal data is a challenging task in machine learning for healthcare that can provide clinicians with better tools for diagnosis and monitoring of disease. Existing DPM algorithms neglect temporal dependencies among measurements and make parametric assumptions about biomarker trajectories. In addition, they do not model multiple biomarkers jointly and need to align subjects' trajectories. In this paper, recurrent neural networks (RNNs) are utilized to address these issues. However, in many cases, longitudinal cohorts contain incomplete data, which hinders the application of standard RNNs and requires a pre-processing step such as imputation of the missing values. We, therefore, propose a generalized training rule for the most widely used RNN architecture, long short-term memory (LSTM) networks, that can handle missing values in both target and predictor variables. This algorithm is applied for modeling the progression of Alzheimer's disease (AD) using magnetic resonance imaging (MRI) biomarkers. The results show that the proposed LSTM algorithm achieves a lower mean absolute error for prediction of measurements across all considered MRI biomarkers compared to using standard LSTM networks with data imputation or using a regression-based DPM method. Moreover, applying linear discriminant analysis to the biomarkers' values predicted by the proposed algorithm results in a larger area under the receiver operating characteristic curve (AUC) for clinical diagnosis of AD compared to the same alternatives, and the AUC is comparable to state-of-the-art AUC's from a recent cross-sectional medical image classification challenge. This paper shows that built-in handling of missing values in LSTM network training paves the way for application of RNNs in disease progression modeling.

AB - Disease progression modeling (DPM) using longitudinal data is a challenging task in machine learning for healthcare that can provide clinicians with better tools for diagnosis and monitoring of disease. Existing DPM algorithms neglect temporal dependencies among measurements and make parametric assumptions about biomarker trajectories. In addition, they do not model multiple biomarkers jointly and need to align subjects' trajectories. In this paper, recurrent neural networks (RNNs) are utilized to address these issues. However, in many cases, longitudinal cohorts contain incomplete data, which hinders the application of standard RNNs and requires a pre-processing step such as imputation of the missing values. We, therefore, propose a generalized training rule for the most widely used RNN architecture, long short-term memory (LSTM) networks, that can handle missing values in both target and predictor variables. This algorithm is applied for modeling the progression of Alzheimer's disease (AD) using magnetic resonance imaging (MRI) biomarkers. The results show that the proposed LSTM algorithm achieves a lower mean absolute error for prediction of measurements across all considered MRI biomarkers compared to using standard LSTM networks with data imputation or using a regression-based DPM method. Moreover, applying linear discriminant analysis to the biomarkers' values predicted by the proposed algorithm results in a larger area under the receiver operating characteristic curve (AUC) for clinical diagnosis of AD compared to the same alternatives, and the AUC is comparable to state-of-the-art AUC's from a recent cross-sectional medical image classification challenge. This paper shows that built-in handling of missing values in LSTM network training paves the way for application of RNNs in disease progression modeling.

M3 - Paper

T2 - 1st Conference on Medical Imaging with Deep Learning (MIDL 2018)

Y2 - 4 July 2018 through 6 July 2018

ER -

ID: 199022614