Deep Learning to Simulate Contrast-enhanced Breast MRI of Invasive Breast Cancer.

Background There is increasing interest in noncontrast breast MRI alternatives for tumor visualization to increase the accessibility of breast MRI. Purpose To evaluate the feasibility and accuracy of generating simulated contrast-enhanced T1-weighted breast MRI scans from precontrast MRI sequences in biopsy-proven invasive breast cancer with use of deep learning. Materials and Methods Women with invasive breast cancer and a contrast-enhanced breast MRI examination that was performed for initial evaluation of the extent of disease between January 2015 and December 2019 at a single academic institution were retrospectively identified. A three-dimensional, fully convolutional deep neural network simulated contrast-enhanced T1-weighted breast MRI scans from five precontrast sequences (T1-weighted non-fat-suppressed [FS], T1-weighted FS, T2-weighted FS, apparent diffusion coefficient, and diffusion-weighted imaging). For qualitative assessment, four breast radiologists (with 3-15 years of experience) blinded to whether the method of contrast was real or simulated assessed image quality (excellent, acceptable, good, poor, or unacceptable), presence of tumor enhancement, and maximum index mass size by using 22 pairs of real and simulated contrast-enhanced MRI scans. Quantitative comparison was performed using whole-breast similarity and error metrics and Dice coefficient analysis of enhancing tumor overlap. Results Ninety-six MRI examinations in 96 women (mean age, 52 years ± 12 [SD]) were evaluated. The readers assessed all simulated MRI scans as having the appearance of a real MRI scan with tumor enhancement. Index mass sizes on real and simulated MRI scans demonstrated good to excellent agreement (intraclass correlation coefficient, 0.73-0.86; P < .001) without significant differences (mean differences, -0.8 to 0.8 mm; P = .36-.80). Almost all simulated MRI scans (84 of 88 [95%]) were considered of diagnostic quality (ratings of excellent, acceptable, or good). Quantitative analysis demonstrated strong similarity (structural similarity index, 0.88 ± 0.05), low voxel-wise error (symmetric mean absolute percent error, 3.26%), and Dice coefficient of enhancing tumor overlap of 0.75 ± 0.25. Conclusion It is feasible to generate simulated contrast-enhanced breast MRI scans with use of deep learning. Simulated and real contrast-enhanced MRI scans demonstrated comparable tumor sizes, areas of tumor enhancement, and image quality without significant qualitative or quantitative differences. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Slanetz in this issue. An earlier incorrect version appeared online. This article was corrected on January 17, 2023.

Correlating Quantitative MRI-based Apparent Diffusion Coefficient Metrics with 24-month Neurodevelopmental Outcomes in Neonates from the HEAL Trial.

Background Multiple qualitative scoring systems have been created to capture the imaging severity of hypoxic ischemic brain injury. Purpose To evaluate quantitative volumes of acute brain injury at MRI in neonates with hypoxic ischemic brain injury and correlate these findings with 24-month neurodevelopmental outcomes and qualitative brain injury scoring by radiologists. Materials and Methods In this secondary analysis, brain diffusion-weighted MRI data from neonates in the High-dose Erythropoietin for Asphyxia and Encephalopathy trial, which recruited participants between January 2017 and October 2019, were analyzed. Volume of acute brain injury, defined as brain with apparent diffusion coefficient (ADC) less than 800 × 10-6 mm2/sec, was automatically computed across the whole brain and within the thalami and white matter. Outcomes of death and neurodevelopmental impairment (NDI) were recorded at 24-month follow-up. Associations between the presence and volume (in milliliters) of acute brain injury with 24-month outcomes were evaluated using multiple logistic regression. The correlation between quantitative acute brain injury volume and qualitative MRI scores was assessed using the Kendall tau-b test. Results A total of 416 neonates had available MRI data (mean gestational age, 39.1 weeks ± 1.4 [SD]; 235 male) and 113 (27%) showed evidence of acute brain injury at MRI. Of the 387 participants with 24-month follow-up data, 185 (48%) died or had any NDI. Volume of acute injury greater than 1 mL (odds ratio [OR], 13.9 [95% CI: 5.93, 32.45]; P < .001) and presence of any acute injury in the brain (OR, 4.5 [95% CI: 2.6, 7.8]; P < .001) were associated with increased odds of death or any NDI. Quantitative whole-brain acute injury volume was strongly associated with radiologists' qualitative scoring of diffusion-weighted images (Kendall tau-b = 0.56; P < .001). Conclusion Automated quantitative volume of brain injury is associated with death, moderate to severe NDI, and cerebral palsy in neonates with hypoxic ischemic encephalopathy and correlated well with qualitative MRI scoring of acute brain injury. Clinical trial registration no. NCT02811263 © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Huisman in this issue.

Reducing Gadolinium Contrast With Artificial Intelligence.

Gadolinium contrast is an important agent in magnetic resonance imaging (MRI), particularly in neuroimaging where it can help identify blood-brain barrier breakdown from an inflammatory, infectious, or neoplastic process. However, gadolinium contrast has several drawbacks, including nephrogenic systemic fibrosis, gadolinium deposition in the brain and bones, and allergic-like reactions. As computer hardware and technology continues to evolve, machine learning has become a possible solution for eliminating or reducing the dose of gadolinium contrast. This review summarizes the clinical uses of gadolinium contrast, the risks of gadolinium contrast, and state-of-the-art machine learning methods that have been applied to reduce or eliminate gadolinium contrast administration, as well as their current limitations, with a focus on neuroimaging applications. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 1.

Association of partial T2-FLAIR mismatch sign and isocitrate dehydrogenase mutation in WHO grade 4 gliomas: results from the ReSPOND consortium.

PURPOSE: While the T2-FLAIR mismatch sign is highly specific for isocitrate dehydrogenase (IDH)-mutant, 1p/19q-noncodeleted astrocytomas among lower-grade gliomas, its utility in WHO grade 4 gliomas is not well-studied. We derived the partial T2-FLAIR mismatch sign as an imaging biomarker for IDH mutation in WHO grade 4 gliomas. METHODS: Preoperative MRI scans of adult WHO grade 4 glioma patients (n = 2165) from the multi-institutional ReSPOND (Radiomics Signatures for PrecisiON Diagnostics) consortium were analyzed. Diagnostic performance of the partial T2-FLAIR mismatch sign was evaluated. Subset analyses were performed to assess associations of imaging markers with overall survival (OS). RESULTS: One hundred twenty-one (5.6%) of 2165 grade 4 gliomas were IDH-mutant. Partial T2-FLAIR mismatch was present in 40 (1.8%) cases, 32 of which were IDH-mutant, yielding 26.4% sensitivity, 99.6% specificity, 80.0% positive predictive value, and 95.8% negative predictive value. Multivariate logistic regression demonstrated IDH mutation was significantly associated with partial T2-FLAIR mismatch (odds ratio [OR] 5.715, 95% CI [1.896, 17.221], p = 0.002), younger age (OR 0.911 [0.895, 0.927], p < 0.001), tumor centered in frontal lobe (OR 3.842, [2.361, 6.251], p < 0.001), absence of multicentricity (OR 0.173, [0.049, 0.612], p = 0.007), and presence of cystic (OR 6.596, [3.023, 14.391], p < 0.001) or non-enhancing solid components (OR 6.069, [3.371, 10.928], p < 0.001). Multivariate Cox analysis demonstrated cystic components (p = 0.024) and non-enhancing solid components (p = 0.003) were associated with longer OS, while older age (p < 0.001), frontal lobe center (p = 0.008), multifocality (p < 0.001), and multicentricity (p < 0.001) were associated with shorter OS. CONCLUSION: Partial T2-FLAIR mismatch sign is highly specific for IDH mutation in WHO grade 4 gliomas.

A Deep Learning Approach for Automated Bone Removal from Computed Tomography Angiography of the Brain.

Advanced visualization techniques such as maximum intensity projection (MIP) and volume rendering (VR) are useful for evaluating neurovascular anatomy on CT angiography (CTA) of the brain; however, interference from surrounding osseous anatomy is common. Existing methods for removing bone from CTA images are limited in scope and/or accuracy, particularly at the skull base. We present a new brain CTA bone removal tool, which addresses many of these limitations. A deep convolutional neural network was designed and trained for bone removal using 72 brain CTAs. The model was tested on 15 CTAs from the same data source and 17 CTAs from an independent external dataset. Bone removal accuracy was assessed quantitatively, by comparing automated segmentation results to manual segmentations, and qualitatively by evaluating VR visualization of the carotid siphons compared to an existing method for automated bone removal. Average Dice overlap between automated and manual segmentations from the internal and external test datasets were 0.986 and 0.979 respectively. This was superior compared to a publicly available noncontrast head CT bone removal algorithm which had a Dice overlap of 0.947 (internal dataset) and 0.938 (external dataset). Our algorithm yielded better VR visualization of the carotid siphons than the publicly available bone removal tool in 14 out of 15 CTAs (93%, chi-square statistic of 22.5, p-value of < 0.00001) from the internal test dataset and 15 out of 17 CTAs (88%, chi-square statistic of 23.1, p-value of < 0.00001) from the external test dataset. Bone removal allowed subjectively superior MIP and VR visualization of vascular anatomy/pathology. The proposed brain CTA bone removal algorithm is rapid, accurate, and allows superior visualization of vascular anatomy and pathology compared to other available techniques and was validated on an independent external dataset.

A high-resolution interactive atlas of the human brainstem using magnetic resonance imaging.

Conventional atlases of the human brainstem are limited by the inflexible, sparsely-sampled, two-dimensional nature of histology, or the low spatial resolution of conventional magnetic resonance imaging (MRI). Postmortem high-resolution MRI circumvents the challenges associated with both modalities. A single human brainstem specimen extending from the rostral diencephalon through the caudal medulla was prepared for imaging after the brain was removed from a 65-year-old male within 24 h of death. The specimen was formalin-fixed for two weeks, then rehydrated and placed in a custom-made MRI compatible tube and immersed in liquid fluorocarbon. MRI was performed in a 7-Tesla scanner with 120 unique diffusion directions. Acquisition time for anatomic and diffusion images were 14 h and 208 h, respectively. Segmentation was performed manually. Deterministic fiber tractography was done using strategically chosen regions of interest and avoidance, with manual editing using expert knowledge of human neuroanatomy. Anatomic and diffusion images were rendered with isotropic resolutions of 50 µm and 200 µm, respectively. Ninety different structures were segmented and labeled, and 11 different fiber bundles were rendered with tractography. The complete atlas is available online for interactive use at https://www.civmvoxport.vm.duke.edu/voxbase/login.php?return_url=%2Fvoxbase%2F. This atlas presents multiple contrasting datasets and selected tract reconstruction with unprecedented resolution for MR imaging of the human brainstem. There are immediate applications in neuroanatomical education, with the potential to serve future applications for neuroanatomical research and enhanced neurosurgical planning through "safe" zones of entry into the human brainstem.

The regional pattern of abnormal cerebrovascular reactivity in HIV-infected, virally suppressed women.

The purpose of this study was to assess whole brain and regional patterns of cerebrovascular reactivity (CVR) abnormalities in HIV-infected women using quantitative whole brain arterial spin labeling (ASL). We hypothesized that HIV-infected women would demonstrate decreased regional brain CVR despite viral suppression. This cross-sectional study recruited subjects from the Bay Area Women's Interagency Health Study (WIHS)-a cohort study designed to investigate the progression of HIV disease in women. In addition to conventional noncontrast cerebral MRI sequences, perfusion imaging was performed before and after the administration of intravenous acetazolamide. CVR was measured by comparing quantitative ASL brain perfusion before and after administration of intravenous acetazolamide. In order to validate and corroborate ASL-based whole brain and regional perfusion, phase-contrast (PC) imaging was also performed through the major neck vessels. FLAIR and susceptibility weighted sequences were performed to assess for white matter injury and microbleeds, respectively. Ten HIV-infected women and seven uninfected, age-matched controls were evaluated. Significant group differences were present in whole brain and regional CVR between HIV-infected and uninfected women. These regional differences were significant in the frontal lobe and basal ganglia. CVR measurements were not significantly impacted by the degree of white matter signal abnormality or presence of microbleeds. Despite complete viral suppression, dysfunction of the neurovascular unit persists in the HIV population. Given the lack of association between CVR and traditional imaging markers of small vessel disease, CVR quantification may provide an early biomarker of pre-morbid vascular disease.

Structural mapping with fiber tractography of the human cuneate fasciculus at microscopic resolution in cervical region.

Human spinal white matter tract anatomy has been mapped using post mortem histological information with the help of molecular tracing studies in animal models. This study used 7 Tesla diffusion MR tractography on a human cadaver that was harvested 24 hours post mortem to evaluate cuneate fasciculus anatomy in cervical spinal cord. Based on this method, for the first time much more nuanced tractographic anatomy was used to investigate possible new routes for cuneate fasciculus in the posterior and lateral funiculus. Additionally, current molecular tracing studies were reviewed, and confirmatory data was presented along with our radiological results. Both studies confirm that upon entry to the spinal cord, upper cervical level tracts (C1-2-3) travel inside lateral funiculus and lower level tracts travel medially inside the posterior funiculus after entry at posterolateral sulcus which is different than traditional knowledge of having cuneate fasciculus tracts concentrated in the lateral part of posterior funiculus.

A Diffusion MRI Tractography Connectome of the Mouse Brain and Comparison with Neuronal Tracer Data.

Interest in structural brain connectivity has grown with the understanding that abnormal neural connections may play a role in neurologic and psychiatric diseases. Small animal connectivity mapping techniques are particularly important for identifying aberrant connectivity in disease models. Diffusion magnetic resonance imaging tractography can provide nondestructive, 3D, brain-wide connectivity maps, but has historically been limited by low spatial resolution, low signal-to-noise ratio, and the difficulty in estimating multiple fiber orientations within a single image voxel. Small animal diffusion tractography can be substantially improved through the combination of ex vivo MRI with exogenous contrast agents, advanced diffusion acquisition and reconstruction techniques, and probabilistic fiber tracking. Here, we present a comprehensive, probabilistic tractography connectome of the mouse brain at microscopic resolution, and a comparison of these data with a neuronal tracer-based connectivity data from the Allen Brain Atlas. This work serves as a reference database for future tractography studies in the mouse brain, and demonstrates the fundamental differences between tractography and neuronal tracer data.

Addendum to "Waxholm Space atlas of the Sprague Dawley rat brain" [NeuroImage 97 (2014) 374-386].

The main focus of our original article was to describe the anatomical delineations constituting the first version of the WHS Sprague Dawley atlas, apply the Waxholm Space coordinate system, and publish the associated MRI/DTI template and segmentation volume in their original format. To increase usability of the dataset, we have recently shared an updated version of the volumetric image material (v1.01). The aims of this addendum are to inform about the improvements in the updated dataset, in particular related to navigation in the WHS coordinate system, and provide guidance for transforming coordinates acquired in the first version of the atlas.

A diffusion tensor MRI atlas of the postmortem rhesus macaque brain.

The rhesus macaque (Macaca mulatta) is the most widely used nonhuman primate for modeling the structure and function of the brain. Brain atlases, and particularly those based on magnetic resonance imaging (MRI), have become important tools for understanding normal brain structure, and for identifying structural abnormalities resulting from disease states, exposures, and/or aging. Diffusion tensor imaging (DTI)-based MRI brain atlases are widely used in both human and macaque brain imaging studies because of the unique contrasts, quantitative diffusion metrics, and diffusion tractography that they can provide. Previous MRI and DTI atlases of the rhesus brain have been limited by low contrast and/or low spatial resolution imaging. Here we present a microscopic resolution MRI/DTI atlas of the rhesus brain based on 10 postmortem brain specimens. The atlas includes both structural MRI and DTI image data, a detailed three-dimensional segmentation of 241 anatomic structures, diffusion tractography, cortical thickness estimates, and maps of anatomic variability among atlas specimens. This atlas incorporates many useful features from previous work, including anatomic label nomenclature and ontology, data orientation, and stereotaxic reference frame, and further extends prior analyses with the inclusion of high-resolution multi-contrast image data.

Postmortem diffusion MRI of the human brainstem and thalamus for deep brain stimulator electrode localization.

Deep brain stimulation (DBS) is an established surgical therapy for medically refractory tremor disorders including essential tremor (ET) and is currently under investigation for use in a variety of other neurologic and psychiatric disorders. There is growing evidence that the anti-tremor effects of DBS for ET are directly related to modulation of the dentatorubrothalamic tract (DRT), a white matter pathway that connects the cerebellum, red nucleus, and ventral intermediate nucleus of the thalamus. Emerging white matter targets for DBS, like the DRT, will require improved three-dimensional (3D) reference maps of deep brain anatomy and structural connectivity for accurate electrode targeting. High-resolution diffusion MRI of postmortem brain specimens can provide detailed volumetric images of important deep brain nuclei and 3D reconstructions of white matter pathways with probabilistic tractography techniques. We present a high spatial and angular resolution diffusion MRI template of the postmortem human brainstem and thalamus with 3D reconstructions of the nuclei and white matter tracts involved in ET circuitry. We demonstrate registration of these data to in vivo, clinical images from patients receiving DBS therapy, and correlate electrode proximity to tractography of the DRT with improvement of ET symptoms.

Waxholm Space atlas of the Sprague Dawley rat brain.

Three-dimensional digital brain atlases represent an important new generation of neuroinformatics tools for understanding complex brain anatomy, assigning location to experimental data, and planning of experiments. We have acquired a microscopic resolution isotropic MRI and DTI atlasing template for the Sprague Dawley rat brain with 39 µm isotropic voxels for the MRI volume and 78 µm isotropic voxels for the DTI. Building on this template, we have delineated 76 major anatomical structures in the brain. Delineation criteria are provided for each structure. We have applied a spatial reference system based on internal brain landmarks according to the Waxholm Space standard, previously developed for the mouse brain, and furthermore connected this spatial reference system to the widely used stereotaxic coordinate system by identifying cranial sutures and related stereotaxic landmarks in the template using contrast given by the active staining technique applied to the tissue. With the release of the present atlasing template and anatomical delineations, we provide a new tool for spatial orientation analysis of neuroanatomical location, and planning and guidance of experimental procedures in the rat brain. The use of Waxholm Space and related infrastructures will connect the atlas to interoperable resources and services for multi-level data integration and analysis across reference spaces.

Investigating the tradeoffs between spatial resolution and diffusion sampling for brain mapping with diffusion tractography: time well spent?

Interest in mapping white matter pathways in the brain has peaked with the recognition that altered brain connectivity may contribute to a variety of neurologic and psychiatric diseases. Diffusion tractography has emerged as a popular method for postmortem brain mapping initiatives, including the ex-vivo component of the human connectome project, yet it remains unclear to what extent computer-generated tracks fully reflect the actual underlying anatomy. Of particular concern is the fact that diffusion tractography results vary widely depending on the choice of acquisition protocol. The two major acquisition variables that consume scan time, spatial resolution, and diffusion sampling, can each have profound effects on the resulting tractography. In this analysis, we determined the effects of the temporal tradeoff between spatial resolution and diffusion sampling on tractography in the ex-vivo rhesus macaque brain, a close primate model for the human brain. We used the wealth of autoradiography-based connectivity data available for the rhesus macaque brain to assess the anatomic accuracy of six time-matched diffusion acquisition protocols with varying balance between spatial and diffusion sampling. We show that tractography results vary greatly, even when the subject and the total acquisition time are held constant. Further, we found that focusing on either spatial resolution or diffusion sampling at the expense of the other is counterproductive. A balanced consideration of both sampling domains produces the most anatomically accurate and consistent results.

A high-resolution cardiovascular magnetic resonance diffusion tensor map from ex-vivo C57BL/6 murine hearts.

BACKGROUND: The complex cardiac fiber structural organization and spatial arrangement of cardiomyocytes in laminar sheetlets contributes greatly to cardiac functional and contractile ejection patterns. This study presents the first comprehensive, ultra-high resolution, fully quantitative statistical tensor map of the fixed murine heart at isotropic resolution of 43 µm using diffusion tensor (DT) cardiovascular magnetic resonance (CMR). METHODS: Imaging was completed in approximately 12 hours using a six-directional encoding scheme, in five ex vivo healthy C57BL/6 mouse hearts. The tensor map constructed from this data provides an average description of the murine fiber architecture visualized with fiber tractography, and its population variability, using the latest advances in image tensor analysis and statistics. RESULTS: Results show that non-normalized cardiac tensor maps are associated with mean fractional anisotropy of 0.25 ± 0.07 and mean diffusivity of 8.9 ± 1.6 × 10⁻4mm²/s. Moreover, average mid-ventricular helical angle distributions ranged between -41 ± 3° and +52 ± 5° and were highly correlated with transmural depth, in agreement with prior published results in humans and canines. Calculated variabilities of local myocyte orientations were 2.0° and 1.4°. Laminar sheet orientation variability was found to be less stable at 2.6°. Despite such variations, the murine heart seems to be highly structured, particularly when compared to canines and humans. CONCLUSIONS: This tensor map has the potential to yield an accurate mean representation and identification of common or unique features of the cardiac myocyte architecture, to establish a baseline standard reference of DTI indices, and to improve detection of biomarkers, especially in pathological states or post-transgenetic modifications.

Segmentation of the canine corpus callosum using diffusion-tensor imaging tractography.

OBJECTIVE: We set out to determine functional white matter (WM) connections passing through the canine corpus callosum; these WM connections would be useful for subsequent studies of canine brains that serve as models for human WM pathway disease. Based on prior studies, we anticipated that the anterior corpus callosum would send projections to the anterior cerebral cortex whereas progressively posterior segments would send projections to more posterior cortex. MATERIALS AND METHODS: A postmortem canine brain was imaged using a 7-T MRI system producing 100-µm-isotropic-resolution diffusion-tensor imaging analyzed by tractography. Using regions of interest (ROIs) within cortical locations, which were confirmed by a Nissl stain that identified distinct cortical architecture, we successfully identified six important WM pathways. We also compared fractional anisotropy (FA), apparent diffusion coefficient (ADC), radial diffusivity, and axial diffusivity in tracts passing through the genu and splenium. RESULTS: Callosal fibers were organized on the basis of cortical destination (e.g., fibers from the genu project to the frontal cortex). Histologic results identified the motor cortex on the basis of cytoarchitectonic criteria that allowed placement of ROIs to discriminate between frontal and parietal lobes. We also identified cytoarchitecture typical of the orbital frontal, anterior frontal, and occipital regions and placed ROIs accordingly. FA, ADC, radial diffusivity, and axial diffusivity values were all higher in posterior corpus callosum fiber tracts. CONCLUSION: Using six cortical ROIs, we identified six major WM tracts that reflect major functional divisions of the cerebral hemispheres, and we derived quantitative values that can be used for study of canine models of human WM pathologic states.

Elevated meningioma risk among individuals who are Non-Hispanic Black is strongest for grade 2-3 tumors and synergistically modified by male sex.

Background: Meningioma risk factors include older age, female sex, and African-American race. There are limited data exploring how meningioma risk in African-Americans varies across the lifespan, interacts with sex, and differs by tumor grade. Methods: The Central Brain Tumor Registry of the United States (CBTRUS) is a population-based registry covering the entire U.S. population. Meningioma diagnoses from 2004-2019 were used to calculate incidence rate ratios (IRRs) for non-Hispanic Black individuals (NHB) compared to non-Hispanic white individuals (NHW) across 10-year age intervals, and stratified by sex and by WHO tumor grade. Results: 53,890 NHB individuals and 322,373 NHW individuals with an intracranial meningioma diagnosis were included in analyses. Beginning in young adulthood, the NHB-to-NHW IRR was elevated for both grade 1 and grade 2/3 tumors. The IRR peaked in the seventh decade of life regardless of grade, and was higher for grade 2/3 tumors (IRR=1.57; 95% CI: 1.46-1.69) than grade 1 tumors (IRR=1.27; 95% CI: 1.25-1.30) in this age group. The NHB-to-NHW IRR was elevated in females (IRR=1.17; 95% CI: 1.16-1.18) and further elevated in males (IRR=1.28; 95% CI: 1.26-1.30), revealing synergistic interaction between NHB race/ethnicity and male sex (P Interaction =0.001). Conclusions: Relative to NHW individuals, NHB individuals are at elevated risk of meningioma from young adulthood through old age. NHB race/ethnicity conferred higher risk of meningioma among men than women, and higher risk of developing WHO grade 2/3 tumors. Results identify meningioma as a significant source of racial disparities in neuro-oncology and may help to improve preoperative predictions of meningioma grade.

Automated neonatal nnU-Net brain MRI extractor trained on a large multi-institutional dataset.

Brain extraction, or skull-stripping, is an essential data preprocessing step for machine learning approaches to brain MRI analysis. Currently, there are limited extraction algorithms for the neonatal brain. We aim to adapt an established deep learning algorithm for the automatic segmentation of neonatal brains from MRI, trained on a large multi-institutional dataset for improved generalizability across image acquisition parameters. Our model, ANUBEX (automated neonatal nnU-Net brain MRI extractor), was designed using nnU-Net and was trained on a subset of participants (N = 433) enrolled in the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) study. We compared the performance of our model to five publicly available models (BET, BSE, CABINET, iBEATv2, ROBEX) across conventional and machine learning methods, tested on two public datasets (NIH and dHCP). We found that our model had a significantly higher Dice score on the aggregate of both data sets and comparable or significantly higher Dice scores on the NIH (low-resolution) and dHCP (high-resolution) datasets independently. ANUBEX performs similarly when trained on sequence-agnostic or motion-degraded MRI, but slightly worse on preterm brains. In conclusion, we created an automatic deep learning-based neonatal brain extraction algorithm that demonstrates accurate performance with both high- and low-resolution MRIs with fast computation time.

The University of California San Francisco Adult Longitudinal Post-Treatment Diffuse Glioma (UCSF-ALPTDG) MRI Dataset.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. The University of California San Francisco Adult Longitudinal Post-Treatment Diffuse Glioma (UCSF-ALPTDG) MRI dataset is a publicly available annotated dataset featuring multimodal brain MRIs from 298 patients with diffuse gliomas taken at two consecutive follow-ups (596 scans total), with corresponding clinical history and expert voxelwise annotations. ©RSNA, 2024.