SynthSR: A public AI tool to turn heterogeneous clinical brain scans into high-resolution T1-weighted images for 3D morphometry

Iglesias, Juan Eugenio; Billot, Benjamin; Balbastre, Yaël; Magdamo, Colin; Arnold, Steven E.; Das, Sudeshna; Edlow, Brian L.; Alexander, Daniel C.; Golland, Polina; Fischl, Bruce

doi:10.1126/sciadv.add3607

Cited by 51 publications

(44 citation statements)

References 86 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to make the neural network agnostic to the orientation (axial, coronal, sagittal), MRI modality, and resolution of the inputs (slice thickness, slice spacing), we adopt a domain randomization approach that we have successfully used in other brain MRI analysis problems (e.g., 26 , 54 ). The key idea is that, by randomizing the orientation, resolution, and MR contrast at every iteration during training, the CNN learns features that are agnostic to MR contrast and resolution.…”

Section: Methodsmentioning

confidence: 99%

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

Iglesias

2023

Sci Rep

View full text Add to dashboard Cite

Volumetric registration of brain MRI is routinely used in human neuroimaging, e.g., to align different MRI modalities, to measure change in longitudinal analysis, to map an individual to a template, or in registration-based segmentation. Classical registration techniques based on numerical optimization have been very successful in this domain, and are implemented in widespread software suites like ANTs, Elastix, NiftyReg, or DARTEL. Over the last 7–8 years, learning-based techniques have emerged, which have a number of advantages like high computational efficiency, potential for higher accuracy, easy integration of supervision, and the ability to be part of a meta-architectures. However, their adoption in neuroimaging pipelines has so far been almost inexistent. Reasons include: lack of robustness to changes in MRI modality and resolution; lack of robust affine registration modules; lack of (guaranteed) symmetry; and, at a more practical level, the requirement of deep learning expertise that may be lacking at neuroimaging research sites. Here, we present EasyReg, an open-source, learning-based registration tool that can be easily used from the command line without any deep learning expertise or specific hardware. EasyReg combines the features of classical registration tools, the capabilities of modern deep learning methods, and the robustness to changes in MRI modality and resolution provided by our recent work in domain randomization. As a result, EasyReg is: fast; symmetric; diffeomorphic (and thus invertible); agnostic to MRI modality and resolution; compatible with affine and nonlinear registration; and does not require any preprocessing or parameter tuning. We present results on challenging registration tasks, showing that EasyReg is as accurate as classical methods when registering 1 mm isotropic scans within MRI modality, but much more accurate across modalities and resolutions. EasyReg is publicly available as part of FreeSurfer; see https://surfer.nmr.mgh.harvard.edu/fswiki/EasyReg.

show abstract

Section: Methodsmentioning

confidence: 99%

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

Iglesias

2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…While adapting the 3D reconstruction (Equation 1) is straightforward, the U-net will need additional image synthesis and manual labeling efforts – particularly if one wishes to include new regions, such as brainstem nuclei. Additional future analyses will include: correlating the segmentation-derived volumes with clinical scores, disease subtypes, and disease duration; using techniques like SynthSR ( Iglesias et al, 2023 ) to improve the resolution of the reconstructed volumes; exploring nonlinear deformation models for the 3D reconstruction; fully automatizing tissue segmentation from the background using neural networks; and extending the tools to 3D analysis of histological sections.…”

Section: Discussionmentioning

confidence: 99%

Machine learning of dissection photographs and surface scanning for quantitative 3D neuropathology

Gazula,

Tregidgo,

Billot

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Dissection photographs are routinely acquired in brain banks but seldom used for quantitative analysis. We present a publicly available, open-source toolbox of computational methods for 3D analysis of photographs of dissected slices of human brains. The toolbox includes methods for: (i) Geometric preprocessing of the photographs; (ii) 3D volumetric reconstruction of a human brain or hemisphere from the photographs and, optionally, a surface scan; and (iii) automatic 3D segmentation of nine bilateral brain regions using a novel machine learning method ("Photo-SynthSeg:), which is capable of "interpolating" the 3D segmentation across the gaps between photographs with large spacing in between. Our tools can be used for quantitative brain analysis at the macroscopic level (e.g., volumetry), as a substitute for ex vivo magnetic resonance imaging (MRI) -- which requires access to an MRI scanner and ex vivo scanning expertise. Our tools are also helpful for computationally guided dissection of brain areas that are difficult to identify from the slices, e.g., defined by functional atlases. Finally, our tools also provide a spatial mapping between the a 3D MRI scan (when available) and the 2D slices, which enables accurate neuropathological-neuroimaging correlation. We evaluate our toolbox on a synthetic dataset and on dissection photographs from two NIH Alzheimer's Disease Research Centers (ADRC). Our methodology yields accurate 3D reconstruction and automated segmentations that are highly correlated to those obtained from ex vivo MRI, and that can detect expected differences between brains with Alzheimer's disease and controls. The tools are available as part of our widespread neuroimaging suite "FreeSurfer" (https://surfer.nmr.mgh.harvard.edu/fswiki/PhotoTools).

show abstract

“…The proposed pipeline flow-chart is displayed in Figure 1 . It fills a cropped T1w MRI with voxels synthesized from a T2w sMRI using FreeSurfer’s ‘ SynthSR ’ and ‘ SynthSR Hyperfine ’ tools [16, 17]. Both tools output synthetic T1w MPRAGE 1mm isotropic scans from a given input of any contrast or spatial resolution.…”

Section: Methodsmentioning

confidence: 99%

Crop Filling: a pipeline for repairing memory clinic MRI corrupted by partial brain coverage

Castro-Leal

Whitfield

Praharaju

et al. 2023

Preprint

View full text Add to dashboard Cite

Data-driven solutions offer great promise for improving healthcare. However standard clinical neuroimaging data is subject to real-world imaging artefacts that can render the data unusable for computational research. T1 weighted structural MRI is used in research to obtain volumetric measurements from cortical and subcortical brain regions. However, clinical radiologists often prioritise T2 weighted or FLAIR scans for visual assessment. As such, T1 weighted scans are often acquired but may not be a priority. This can result in artefacts such as partial brain coverage being systematically present in memory clinic data. Here we present a neuroimaging pipeline to ameliorate such situations by filling the missing regions with synthetic data. We validate on artificially cropped scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI), showing that our pipeline largely removes the artefact, improving volumetric biomarker accuracy while also retaining statistical differences between diagnostic groups. We demonstrate utility by achieving diagnostic classification performance comparable to uncorrupted data. This is an important contribution towards moving research from the lab into the real world.

show abstract

SynthSR: A public AI tool to turn heterogeneous clinical brain scans into high-resolution T1-weighted images for 3D morphometry

Cited by 51 publications

References 86 publications

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

Machine learning of dissection photographs and surface scanning for quantitative 3D neuropathology

Crop Filling: a pipeline for repairing memory clinic MRI corrupted by partial brain coverage

Contact Info

Product

Resources

About