Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection

Olfa Ghribi, Gang Li, Weili Lin, Dinggang Shen, Islem Rekik

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The early postnatal developmental period is highly dynamic, where brain connections undergo both growth and pruning processes. Understanding typical brain connectivity evolution would enable us to spot abnormal connectional development patterns. However, this generally requires the acquisition of longitudinal neuroimaging datasets that densely cover the first years of postnatal development. This might not be easily investigated since neonatal follow-up scans are rarely acquired in a clinical setting. Furthermore, waiting for the acquisition of later brain scans would hinder early neurodevelopmental disorder diagnosis. To solve this problem, we unprecedentedly propose a bidirectionally supervised sample selection framework, while leveraging the time-dependency between consecutive observations, for predicting neonatal brain connectome evolution from a single structural magnetic resonance imaging (MRI) acquired around birth. Specifically, we propose to learn how to select the best training samples by supervisedly training a bidirectional ensemble of regressors from the space of pairwise neonatal connectome disparities to their expected prediction scores resulting from using one training connectome to predict another training connectome. The proposed supervised ensemble learning is time-dependent and has a recall memory anchored at the ground truth baseline observation, allowing to progressively pass over previous predictions through the connectome evolution trajectory. We then rank training samples at current timepoint ti-1 based on their expected prediction scores by the ensemble and average their connectomes at follow-up timepoint ti to predict the testing connectome at ti. Our framework significantly outperformed comparison methods in leave-one-out cross-validation.

Original languageEnglish
Title of host publicationPredictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings
EditorsIslem Rekik, Ehsan Adeli, Sang Hyun Park
PublisherSpringer
Pages63-72
Number of pages10
ISBN (Print)9783030322809
DOIs
Publication statusPublished - 2019 Jan 1
Externally publishedYes
Event2nd International Workshop on Predictive Intelligence in Medicine, PRIME 2019, held in conjunction with the 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019 - Shenzhen, China
Duration: 2019 Oct 132019 Oct 13

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11843 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2nd International Workshop on Predictive Intelligence in Medicine, PRIME 2019, held in conjunction with the 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019
CountryChina
CityShenzhen
Period19/10/1319/10/13

Fingerprint

Sample Selection
Magnetic Resonance Imaging
Magnetic resonance
Brain
Connectivity
Imaging techniques
Prediction
Training Samples
Ensemble
Neuroimaging
Ensemble Learning
Predict
Comparison Method
Supervised learning
Supervised Learning
Pruning
Cross-validation
Disorder
Consecutive
Pairwise

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ghribi, O., Li, G., Lin, W., Shen, D., & Rekik, I. (2019). Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection. In I. Rekik, E. Adeli, & S. H. Park (Eds.), Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings (pp. 63-72). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11843 LNCS). Springer. https://doi.org/10.1007/978-3-030-32281-6_7

Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection. / Ghribi, Olfa; Li, Gang; Lin, Weili; Shen, Dinggang; Rekik, Islem.

Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings. ed. / Islem Rekik; Ehsan Adeli; Sang Hyun Park. Springer, 2019. p. 63-72 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11843 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ghribi, O, Li, G, Lin, W, Shen, D & Rekik, I 2019, Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection. in I Rekik, E Adeli & SH Park (eds), Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11843 LNCS, Springer, pp. 63-72, 2nd International Workshop on Predictive Intelligence in Medicine, PRIME 2019, held in conjunction with the 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019, Shenzhen, China, 19/10/13. https://doi.org/10.1007/978-3-030-32281-6_7
Ghribi O, Li G, Lin W, Shen D, Rekik I. Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection. In Rekik I, Adeli E, Park SH, editors, Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings. Springer. 2019. p. 63-72. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-32281-6_7
Ghribi, Olfa ; Li, Gang ; Lin, Weili ; Shen, Dinggang ; Rekik, Islem. / Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection. Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings. editor / Islem Rekik ; Ehsan Adeli ; Sang Hyun Park. Springer, 2019. pp. 63-72 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{7ae1d6ce22354f6bafc512ba17035338,
title = "Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection",
abstract = "The early postnatal developmental period is highly dynamic, where brain connections undergo both growth and pruning processes. Understanding typical brain connectivity evolution would enable us to spot abnormal connectional development patterns. However, this generally requires the acquisition of longitudinal neuroimaging datasets that densely cover the first years of postnatal development. This might not be easily investigated since neonatal follow-up scans are rarely acquired in a clinical setting. Furthermore, waiting for the acquisition of later brain scans would hinder early neurodevelopmental disorder diagnosis. To solve this problem, we unprecedentedly propose a bidirectionally supervised sample selection framework, while leveraging the time-dependency between consecutive observations, for predicting neonatal brain connectome evolution from a single structural magnetic resonance imaging (MRI) acquired around birth. Specifically, we propose to learn how to select the best training samples by supervisedly training a bidirectional ensemble of regressors from the space of pairwise neonatal connectome disparities to their expected prediction scores resulting from using one training connectome to predict another training connectome. The proposed supervised ensemble learning is time-dependent and has a recall memory anchored at the ground truth baseline observation, allowing to progressively pass over previous predictions through the connectome evolution trajectory. We then rank training samples at current timepoint ti-1 based on their expected prediction scores by the ensemble and average their connectomes at follow-up timepoint ti to predict the testing connectome at ti. Our framework significantly outperformed comparison methods in leave-one-out cross-validation.",
author = "Olfa Ghribi and Gang Li and Weili Lin and Dinggang Shen and Islem Rekik",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-32281-6_7",
language = "English",
isbn = "9783030322809",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer",
pages = "63--72",
editor = "Islem Rekik and Ehsan Adeli and Park, {Sang Hyun}",
booktitle = "Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings",

}

TY - GEN

T1 - Progressive infant brain connectivity evolution prediction from neonatal MRI using bidirectionally supervised sample selection

AU - Ghribi, Olfa

AU - Li, Gang

AU - Lin, Weili

AU - Shen, Dinggang

AU - Rekik, Islem

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The early postnatal developmental period is highly dynamic, where brain connections undergo both growth and pruning processes. Understanding typical brain connectivity evolution would enable us to spot abnormal connectional development patterns. However, this generally requires the acquisition of longitudinal neuroimaging datasets that densely cover the first years of postnatal development. This might not be easily investigated since neonatal follow-up scans are rarely acquired in a clinical setting. Furthermore, waiting for the acquisition of later brain scans would hinder early neurodevelopmental disorder diagnosis. To solve this problem, we unprecedentedly propose a bidirectionally supervised sample selection framework, while leveraging the time-dependency between consecutive observations, for predicting neonatal brain connectome evolution from a single structural magnetic resonance imaging (MRI) acquired around birth. Specifically, we propose to learn how to select the best training samples by supervisedly training a bidirectional ensemble of regressors from the space of pairwise neonatal connectome disparities to their expected prediction scores resulting from using one training connectome to predict another training connectome. The proposed supervised ensemble learning is time-dependent and has a recall memory anchored at the ground truth baseline observation, allowing to progressively pass over previous predictions through the connectome evolution trajectory. We then rank training samples at current timepoint ti-1 based on their expected prediction scores by the ensemble and average their connectomes at follow-up timepoint ti to predict the testing connectome at ti. Our framework significantly outperformed comparison methods in leave-one-out cross-validation.

AB - The early postnatal developmental period is highly dynamic, where brain connections undergo both growth and pruning processes. Understanding typical brain connectivity evolution would enable us to spot abnormal connectional development patterns. However, this generally requires the acquisition of longitudinal neuroimaging datasets that densely cover the first years of postnatal development. This might not be easily investigated since neonatal follow-up scans are rarely acquired in a clinical setting. Furthermore, waiting for the acquisition of later brain scans would hinder early neurodevelopmental disorder diagnosis. To solve this problem, we unprecedentedly propose a bidirectionally supervised sample selection framework, while leveraging the time-dependency between consecutive observations, for predicting neonatal brain connectome evolution from a single structural magnetic resonance imaging (MRI) acquired around birth. Specifically, we propose to learn how to select the best training samples by supervisedly training a bidirectional ensemble of regressors from the space of pairwise neonatal connectome disparities to their expected prediction scores resulting from using one training connectome to predict another training connectome. The proposed supervised ensemble learning is time-dependent and has a recall memory anchored at the ground truth baseline observation, allowing to progressively pass over previous predictions through the connectome evolution trajectory. We then rank training samples at current timepoint ti-1 based on their expected prediction scores by the ensemble and average their connectomes at follow-up timepoint ti to predict the testing connectome at ti. Our framework significantly outperformed comparison methods in leave-one-out cross-validation.

UR - http://www.scopus.com/inward/record.url?scp=85075667787&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85075667787&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-32281-6_7

DO - 10.1007/978-3-030-32281-6_7

M3 - Conference contribution

AN - SCOPUS:85075667787

SN - 9783030322809

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 63

EP - 72

BT - Predictive Intelligence in Medicine - 2nd International Workshop, PRIME 2019, Held in Conjunction with MICCAI 2019, Proceedings

A2 - Rekik, Islem

A2 - Adeli, Ehsan

A2 - Park, Sang Hyun

PB - Springer

ER -