Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory

Zuo, Lianrui; Liu, Yihao; He, Yufan; Newsome, Scott D.; Mowry, Ellen M.; Resnick, Susan M.; Prince, Jerry L.; Carass, Aaron

doi:10.1016/j.neuroimage.2021.118569

Cited by 73 publications

(67 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Direct translation of existing harmonisation methods into FL frameworks is non-trivial. Most deep learning methods for harmonisation are based on generative frameworks [3,12,26,27], and, although federated equivalents to GANs and VAEs are being developed [16,25], additional challenges exist for harmonisation approaches that require simultaneous access to source and target data [8]. Additionally, many methods require paired data -not possible with distributed data and unlikely to exist in large multisite studies.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

FedHarmony: Unlearning Scanner Bias with Distributed Data

Dinsdale¹,

Jenkinson²,

Namburete³

2022

Preprint

View full text Add to dashboard Cite

The ability to combine data across scanners and studies is vital for neuroimaging, to increase both statistical power and the representation of biological variability. However, combining datasets across sites leads to two challenges: first, an increase in undesirable non-biological variance due to scanner and acquisition differences -the harmonisation problem -and second, data privacy concerns due to the inherently personal nature of medical imaging data, meaning that sharing them across sites may risk violation of privacy laws. To overcome these restrictions, we propose FedHarmony: a harmonisation framework operating in the federated learning paradigm. We show that to remove the scannerspecific effects, we only need to share the mean and standard deviation of the learned features, helping to protect individual subjects' privacy. We demonstrate our approach across a range of realistic data scenarios, using real multi-site data from the ABIDE dataset, thus showing the potential utility of our method for MRI harmonisation across studies. Our code is available at https://github.com/nkdinsdale/FedHarmony.

show abstract

Section: Introductionmentioning

confidence: 99%

“…Additionally, many methods require paired data -not possible with distributed data and unlikely to exist in large multisite studies. Further, most generative methods are data-hungry [3,27], which casts into doubt whether sufficient data would be available at local sites.…”

Section: Introductionmentioning

confidence: 99%

FedHarmony: Unlearning Scanner Bias with Distributed Data

Dinsdale¹,

Jenkinson²,

Namburete³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The recent development of disentangled representation learning benefits various medical image analysis tasks including segmentation [17,22,10], quality assessment [12,25], domain adaptation [26], and image-to-image translation (I2I) [9,29,30]. The underlying assumption of disentanglement is that a high-dimensional observation x is generated by a latent variable z, where z can be decomposed into independent factors with each factor capturing a certain type of variation of x, i.e., the probability density functions satisfy p(z 1 , z 2 ) = p(z 1 )p(z 2 ) and z = (z 1 , z 2 ) [19].…”

Section: Introductionmentioning

confidence: 99%

“…The underlying assumption of disentanglement is that a high-dimensional observation x is generated by a latent variable z, where z can be decomposed into independent factors with each factor capturing a certain type of variation of x, i.e., the probability density functions satisfy p(z 1 , z 2 ) = p(z 1 )p(z 2 ) and z = (z 1 , z 2 ) [19]. For medical images, it is commonly assumed that z is a composition of contrast (i.e., acquisition-related) and anatomical information of image x [7,9,30,22,17]. While the contrast representations capture specific information about the imaging modality, acquisition parameters, and cohort, the anatomical representations are generally assumed to be invariant to image domains 4 .…”

Section: Introductionmentioning

confidence: 99%

“…It has been shown that the disentangled domain-invariant anatomical representation is a robust input for segmentation [22,7,17,21], and the domain-specific contrast representation provides rich information about image acquisitions. Recombining the disentangled anatomical representation with the desired contrast representation also enables cross-domain I2I [30,20,22].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Disentangling A Single MR Modality

Zuo¹,

Liu²,

Yang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Disentangling anatomical and contrast information from medical images has gained attention recently, demonstrating benefits for various image analysis tasks. Current methods learn disentangled representations using either paired multi-modal images with the same underlying anatomy or auxiliary labels (e.g., manual delineations) to provide inductive bias for disentanglement. However, these requirements could significantly increase the time and cost in data collection and limit the applicability of these methods when such data are not available. Moreover, these methods generally do not guarantee disentanglement. In this paper, we present a novel framework that learns theoretically and practically superior disentanglement from single modality magnetic resonance images. Moreover, we propose a new information-based metric to quantitatively evaluate disentanglement. Comparisons over existing disentangling methods demonstrate that the proposed method achieves superior performance in both disentanglement and cross-domain image-to-image translation tasks.

show abstract

Style transfer generative adversarial networks to harmonize multisite MRI to a single reference image to avoid overcorrection

Liu

Zhu

Maiti

et al. 2023

Human Brain Mapping

View full text Add to dashboard Cite

Recent work within neuroimaging consortia have aimed to identify reproducible, and often subtle, brain signatures of psychiatric or neurological conditions. To allow for high‐powered brain imaging analyses, it is often necessary to pool MR images that were acquired with different protocols across multiple scanners. Current retrospective harmonization techniques have shown promise in removing site‐related image variation. However, most statistical approaches may over‐correct for technical, scanning‐related, variation as they cannot distinguish between confounded image‐acquisition based variability and site‐related population variability. Such statistical methods often require that datasets contain subjects or patient groups with similar clinical or demographic information to isolate the acquisition‐based variability. To overcome this limitation, we consider site‐related magnetic resonance (MR) imaging harmonization as a style transfer problem rather than a domain transfer problem. Using a fully unsupervised deep‐learning framework based on a generative adversarial network (GAN), we show that MR images can be harmonized by inserting the style information encoded from a single reference image, without knowing their site/scanner labels a priori. We trained our model using data from five large‐scale multisite datasets with varied demographics. Results demonstrated that our style‐encoding model can harmonize MR images, and match intensity profiles, without relying on traveling subjects. This model also avoids the need to control for clinical, diagnostic, or demographic information. We highlight the effectiveness of our method for clinical research by comparing extracted cortical and subcortical features, brain‐age estimates, and case–control effect sizes before and after the harmonization. We showed that our harmonization removed the site‐related variances, while preserving the anatomical information and clinical meaningful patterns. We further demonstrated that with a diverse training set, our method successfully harmonized MR images collected from unseen scanners and protocols, suggesting a promising tool for ongoing collaborative studies. Source code is released in USC‐IGC/style_transfer_harmonization (github.com).

show abstract

Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory

Cited by 73 publications

References 21 publications

FedHarmony: Unlearning Scanner Bias with Distributed Data

FedHarmony: Unlearning Scanner Bias with Distributed Data

Disentangling A Single MR Modality

Style transfer generative adversarial networks to harmonize multisite MRI to a single reference image to avoid overcorrection

Contact Info

Product

Resources

About