東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Deep Learning for Anatomical Brain M...

Lijun, An,

Linked to FindBook

Google Book

Amazon

博客來

Deep Learning for Anatomical Brain MRI Harmonization /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Deep Learning for Anatomical Brain MRI Harmonization // An Lijun.
Author:	Lijun, An,
Description:	1 electronic resource (164 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-06, Section: B.
Contained By:	Dissertations Abstracts International86-06B.
Subject:	Tomography. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31777272
ISBN:	9798346775164

Deep Learning for Anatomical Brain MRI Harmonization /
Lijun, An,

Deep Learning for Anatomical Brain MRI Harmonization /An Lijun. - 1 electronic resource (164 pages)

Source: Dissertations Abstracts International, Volume: 86-06, Section: B.

Mega-analyses have significantly advanced neuroimaging research by increasing statistical power, enhancing generalizability, and detecting subtle effects. Mega-analyses work by combining large amounts of data from multiple datasets or sites. Harmonization is a crucial postprocessing step to reduce unwanted inter-site heterogeneity when combining MRI from multiple datasets. Most MRI harmonization algorithms do not explicitly consider predictive downstream applications, which might potentially limit downstream performance. Furthermore, many deep learning harmonization approaches overlook the incorporation of covariates. Since covariate distribution differences are unavoidable across datasets, neglecting covariates during harmonization can inadvertently remove covariate differences instead of reducing undesired inter-site heterogeneity. In this thesis, we focused on addressing the aforementioned research gaps.First, we proposed a goal-specific brain MRI harmonization framework goal-specific conditional variational auto-encoder (gcVAE). MRI harmonization is an ill-posed problem; gcVAE utilizes predictive downstream application performance to regularize the harmonization procedure. Our framework can be flexibly integrated with various deep learning based harmonization models, such as the recently proposed conditional variational autoencoder (cVAE) harmonization model. Using three large-scale datasets with a total of 2787 participants and 10085 anatomical T1 scans, we demonstrated that our approach outperformed existing approaches in preserving relevant biological information while removing site differences.Second, we proposed deep learning models, covariate variational autoencoder (coVAE) and residual batching effects harmonization (ResBat), that account for covariate distribution differences across datasets. coVAE extended cVAE by concatenating covariates with latent representations, while ResBat adopted a residual framework inspired by the classic combating batching effects when combining batches (ComBat) harmonization algorithm. ResBat preserved the effects of covariates using interpretable models and eliminated unwanted inter-site heterogeneity in residuals through a deep learning model. We found that coVAE introduced spurious associations between anatomical MRI and randomly shuffled covariates during harmonization, while ResBat effectively mitigated this false-positive issue. Furthermore, we found that ResBat outperformed a wide suite of classical and deep learning baselines on various evaluation metrics.Overall, this thesis emphasizes the importance of appropriately framing neuroimaging problems under the deep learning paradigm. Ignoring predictive downstream applications, neglecting covariates or improperly incorporating covariates might lead to unsatisfactory outcomes. By addressing these concerns, this thesis contributes to the advancement of neuroimaging harmonization methodology and highlights the importance of adequately modeling these factors in future harmonization studies.

English

ISBN: 9798346775164Subjects--Topical Terms:

836553
Tomography.

Deep Learning for Anatomical Brain MRI Harmonization /
LDR:04315nmm a22004093i 4500 001 2429437
005 20250430100828.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798346775164
035 $a (MiAaPQD)AAI31777272
035 $a AAI31777272
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Lijun, An, $e author. $3 3817284
245 1 0 $a Deep Learning for Anatomical Brain MRI Harmonization / $c An Lijun.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (164 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-06, Section: B.
500 $a Advisors: Thomas, Yeo Boon Thye.
502 $b Ph.D. $c National University of Singapore (Singapore) $d 2024.
520 $a Mega-analyses have significantly advanced neuroimaging research by increasing statistical power, enhancing generalizability, and detecting subtle effects. Mega-analyses work by combining large amounts of data from multiple datasets or sites. Harmonization is a crucial postprocessing step to reduce unwanted inter-site heterogeneity when combining MRI from multiple datasets. Most MRI harmonization algorithms do not explicitly consider predictive downstream applications, which might potentially limit downstream performance. Furthermore, many deep learning harmonization approaches overlook the incorporation of covariates. Since covariate distribution differences are unavoidable across datasets, neglecting covariates during harmonization can inadvertently remove covariate differences instead of reducing undesired inter-site heterogeneity. In this thesis, we focused on addressing the aforementioned research gaps.First, we proposed a goal-specific brain MRI harmonization framework goal-specific conditional variational auto-encoder (gcVAE). MRI harmonization is an ill-posed problem; gcVAE utilizes predictive downstream application performance to regularize the harmonization procedure. Our framework can be flexibly integrated with various deep learning based harmonization models, such as the recently proposed conditional variational autoencoder (cVAE) harmonization model. Using three large-scale datasets with a total of 2787 participants and 10085 anatomical T1 scans, we demonstrated that our approach outperformed existing approaches in preserving relevant biological information while removing site differences.Second, we proposed deep learning models, covariate variational autoencoder (coVAE) and residual batching effects harmonization (ResBat), that account for covariate distribution differences across datasets. coVAE extended cVAE by concatenating covariates with latent representations, while ResBat adopted a residual framework inspired by the classic combating batching effects when combining batches (ComBat) harmonization algorithm. ResBat preserved the effects of covariates using interpretable models and eliminated unwanted inter-site heterogeneity in residuals through a deep learning model. We found that coVAE introduced spurious associations between anatomical MRI and randomly shuffled covariates during harmonization, while ResBat effectively mitigated this false-positive issue. Furthermore, we found that ResBat outperformed a wide suite of classical and deep learning baselines on various evaluation metrics.Overall, this thesis emphasizes the importance of appropriately framing neuroimaging problems under the deep learning paradigm. Ignoring predictive downstream applications, neglecting covariates or improperly incorporating covariates might lead to unsatisfactory outcomes. By addressing these concerns, this thesis contributes to the advancement of neuroimaging harmonization methodology and highlights the importance of adequately modeling these factors in future harmonization studies.
546 $a English
590 $a School code: 1883
650 4 $a Tomography. $3 836553
650 4 $a Neuroimaging. $3 3509452
650 4 $a Alzheimer's disease. $3 588010
650 4 $a Deep learning. $3 3554982
650 4 $a Memory. $3 522110
650 4 $a Brain research. $3 3561789
650 4 $a Scanners. $3 3562953
650 4 $a Magnetic fields. $3 660770
650 4 $a Neurosciences. $3 588700
650 4 $a Neuropsychology. $3 525119
650 4 $a Cognitive ability. $3 3560491
650 4 $a Medical imaging. $3 3172799
650 4 $a Cognition & reasoning. $3 3556293
650 4 $a Magnetic resonance imaging. $3 554355
650 4 $a Dementia. $3 568139
650 4 $a Neural networks. $3 677449
650 4 $a Biomarkers. $3 2205735
650 4 $a Atrophy. $3 3557722
650 4 $a Aging. $3 543123
650 4 $a Cognitive psychology. $3 523881
650 4 $a Electromagnetics. $3 3173223
690 $a 0574
690 $a 0317
690 $a 0493
690 $a 0800
690 $a 0633
690 $a 0607
710 2 $a National University of Singapore (Singapore). $e degree granting institution. $3 3765631
720 1 $a Thomas, Yeo Boon Thye $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-06B.
790 $a 1883
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31777272