Search

This paper demonstrates a self-supervised framework for learning voxel-wise coarse-to-fine representations tailored for dense downstream tasks. Our approach stems from the observation that existing methods for hierarchical representation learning tend to prioritize global features over local features due to inherent architectural bias. To address this challenge, we devise a training strategy that balances the contributions of features from multiple scales, ensuring that the learned representations capture both coarse and fine-grained details. Our strategy incorporates 3-fold improvements: (1) local data augmentations, (2) a hierarchically balanced architecture, and (3) a hybrid contrastive-restorative loss function. We evaluate our method on CT and MRI data and demonstrate that our new approach particularly beneficial for fine-tuning with limited annotated data and consistently outperforms the baseline counterpart in linear evaluation settings., Comment: Accepted for International Symposium on Biomedical Imaging 2024 (ISBI 2024)

The deviation between chronological age and age predicted from neuroimaging data has been identified as a sensitive risk-marker of cross-disorder brain changes, growing into a cornerstone of biological age-research. However, Machine Learning models underlying the field do not consider uncertainty, thereby confounding results with training data density and variability. Also, existing models are commonly based on homogeneous training sets, often not independently validated, and cannot be shared due to data protection issues. Here, we introduce an uncertainty-aware, shareable, and transparent Monte-Carlo Dropout Composite-Quantile-Regression (MCCQR) Neural Network trained on N=10,691 datasets from the German National Cohort. The MCCQR model provides robust, distribution-free uncertainty quantification in high-dimensional neuroimaging data, achieving lower error rates compared to existing models across ten recruitment centers and in three independent validation samples (N=4,004). In two examples, we demonstrate that it prevents spurious associations and increases power to detect accelerated brain-aging. We make the pre-trained model publicly available.

Age prediction based on Magnetic Resonance Imaging (MRI) data of the brain is a biomarker to quantify the progress of brain diseases and aging. Current approaches rely on preparing the data with multiple preprocessing steps, such as registering voxels to a standardized brain atlas, which yields a significant computational overhead, hampers widespread usage and results in the predicted brain-age to be sensitive to preprocessing parameters. Here we describe a 3D Convolutional Neural Network (CNN) based on the ResNet architecture being trained on raw, non-registered T$_ 1$-weighted MRI data of N=10,691 samples from the German National Cohort and additionally applied and validated in N=2,173 samples from three independent studies using transfer learning. For comparison, state-of-the-art models using preprocessed neuroimaging data are trained and validated on the same samples. The 3D CNN using raw neuroimaging data predicts age with a mean average deviation of 2.84 years, outperforming the state-of-the-art brain-age models using preprocessed data. Since our approach is invariant to preprocessing software and parameter choices, it enables faster, more robust and more accurate brain-age modeling.

In patients with multiple sclerosis (MS) non-communicating syringomyelia (NCS) has been described as an incidental finding in case studies and small case series. NCS in MS patients commonly leads to uncertainty particularly as the clinical picture of NCS is variable and surgical therapy may be considered. Up to date little is known about the prevalence and clinical importance of NCS in MS. We report the imaging and clinical characteristics of NCS formations in nine MS patients from a 1 year follow-up study in a representative group of 202 MS (4.5%) patients. Brain and spinal cord MRI was performed as part of a genetic study. NCS did commonly extend the central canal and the cord was slightly distended at the level of the syrinx. The cord and syrinx showed no tendency to change in size or shape over 1 year. Despite thorough search into the clinical history and current clinical status no definite but only minimal indications of symptoms potentially related to the NCS were found. We confirm that NCS may occur in MS patients with spinal cord pathology. It can be a subtle finding without clinical correlates. Syrinx formations are more likely to be a consequence of MS cord pathology than a coincidental finding

BACKGROUND: Brain imaging studies have revealed anatomical anomalies in the brains of individuals with Tourette syndrome (TS). Prefrontal regions have been found to be larger and the corpus callosum (CC) area smaller in children and young adults with TS compared with healthy control subjects, and these anatomical features have been understood to reflect neural plasticity that helps to attenuate the severity of tics. METHOD: CC white matter connectivity, as measured by the Fractional Anisotropy (FA) index from diffusion tensor images, was assessed in 20 clinically well-defined boys with Tourette syndrome and 20 age- and gender-matched controls. RESULTS: The hypothesis that children with TS would show reduced measures of connectivity in CC fibers was confirmed for all subregions of the CC. There was no significant interaction of TS and region. Reductions in FA in CC regions may reflect either fewer interhemispheric fibers or reduced axonal myelination. FA values did not correlate significantly with the severity of tic symptoms. Group differences in measures of connectivity did not seem to be attributable to the presence of comorbid ADHD or OCD, to medication exposure, or group differences in IQ. CONCLUSION: Our findings of a reduced interhemispheral white matter connectivity add to the understanding of neural connectivity and plasticity in the brains of children who have TS.

BACKGROUND: Brain imaging studies have revealed anatomical anomalies in the brains of individuals with Tourette syndrome (TS). Prefrontal regions have been found to be larger and the corpus callosum (CC) area smaller in children and young adults with TS compared with healthy control subjects, and these anatomical features have been understood to reflect neural plasticity that helps to attenuate the severity of tics. METHOD: CC white matter connectivity, as measured by the Fractional Anisotropy (FA) index from diffusion tensor images, was assessed in 20 clinically well-defined boys with Tourette syndrome and 20 age- and gender-matched controls. RESULTS: The hypothesis that children with TS would show reduced measures of connectivity in CC fibers was confirmed for all subregions of the CC. There was no significant interaction of TS and region. Reductions in FA in CC regions may reflect either fewer interhemispheric fibers or reduced axonal myelination. FA values did not correlate significantly with the severity of tic symptoms. Group differences in measures of connectivity did not seem to be attributable to the presence of comorbid ADHD or OCD, to medication exposure, or group differences in IQ. CONCLUSION: Our findings of a reduced interhemispheral white matter connectivity add to the understanding of neural connectivity and plasticity in the brains of children who have TS.

BACKGROUND: Brain imaging studies have revealed anatomical anomalies in the brains of individuals with Tourette syndrome (TS). Prefrontal regions have been found to be larger and the corpus callosum (CC) area smaller in children and young adults with TS compared with healthy control subjects, and these anatomical features have been understood to reflect neural plasticity that helps to attenuate the severity of tics. METHOD: CC white matter connectivity, as measured by the Fractional Anisotropy (FA) index from diffusion tensor images, was assessed in 20 clinically well-defined boys with Tourette syndrome and 20 age- and gender-matched controls. RESULTS: The hypothesis that children with TS would show reduced measures of connectivity in CC fibers was confirmed for all subregions of the CC. There was no significant interaction of TS and region. Reductions in FA in CC regions may reflect either fewer interhemispheric fibers or reduced axonal myelination. FA values did not correlate significantly with the severity of tic symptoms. Group differences in measures of connectivity did not seem to be attributable to the presence of comorbid ADHD or OCD, to medication exposure, or group differences in IQ. CONCLUSION: Our findings of a reduced interhemispheral white matter connectivity add to the understanding of neural connectivity and plasticity in the brains of children who have TS.

The importance of Magnetic Resonance Tomography (MRT) arises from the high frequency of water protons in living tissue, the non-invasiveness of the technique and the straightforward spatial encoding of protons by means of additional locally varying magnetic field gradients. During the last two decades, diffusion-weighted imaging (DWI) has added an essential contribution to clinical MR tomography of structural T1- and T2-weighted imaging and to mapping of physiological processes like perfusion and functional activity. DWI uses the self-diffusion (Brownian motion) of water molecules in an inhomogeneous static magnetic field that can be encoded by locally dependent gradient fields. The key feature and importance of MR-DWI results from the fact that the random translatory motion of molecules scans the microscopic tissue structures far beyond the spatial resolution of MR imaging techniques. Furthermore, diffusion as a physical process is independent of magnetic resonance phenomena, but offers similar advantages like high contrast and spatial resolution. The intent of this dissertation covers projects of sequence development and optimization of clinical DWI applications that illustrate the fast evolving development of DWI in medical MR imaging. One focus is set on the combination of DW echo-planar imaging (EPI) and the fluid-attenuated inversion recovery (FLAIR) prepa¬ration. After technical validation, issues of measurement accuracy and signal-to-noise ratio and their implications on estimated contrast parameters like diffusion anisotropy are discussed. A proposed correction term enables immediate acquisition and comparison of standard DW-EPI and FLAIR-DWI in volunteer and patient studies. A preliminary study on patients with astrocytoma reveals the advantages of FLAIR-prepared DW imaging protocols. The second topic of this dissertation explores the extension of the linear diffusion tensor model to high angular resolution DWI (HARDI). The high complexity of white matter

The importance of Magnetic Resonance Tomography (MRT) arises from the high frequency of water protons in living tissue, the non-invasiveness of the technique and the straightforward spatial encoding of protons by means of additional locally varying magnetic field gradients. During the last two decades, diffusion-weighted imaging (DWI) has added an essential contribution to clinical MR tomography of structural T1- and T2-weighted imaging and to mapping of physiological processes like perfusion and functional activity. DWI uses the self-diffusion (Brownian motion) of water molecules in an inhomogeneous static magnetic field that can be encoded by locally dependent gradient fields. The key feature and importance of MR-DWI results from the fact that the random translatory motion of molecules scans the microscopic tissue structures far beyond the spatial resolution of MR imaging techniques. Furthermore, diffusion as a physical process is independent of magnetic resonance phenomena, but offers similar advantages like high contrast and spatial resolution. The intent of this dissertation covers projects of sequence development and optimization of clinical DWI applications that illustrate the fast evolving development of DWI in medical MR imaging. One focus is set on the combination of DW echo-planar imaging (EPI) and the fluid-attenuated inversion recovery (FLAIR) prepa¬ration. After technical validation, issues of measurement accuracy and signal-to-noise ratio and their implications on estimated contrast parameters like diffusion anisotropy are discussed. A proposed correction term enables immediate acquisition and comparison of standard DW-EPI and FLAIR-DWI in volunteer and patient studies. A preliminary study on patients with astrocytoma reveals the advantages of FLAIR-prepared DW imaging protocols. The second topic of this dissertation explores the extension of the linear diffusion tensor model to high angular resolution DWI (HARDI). The high complexity of white matter

The importance of Magnetic Resonance Tomography (MRT) arises from the high frequency of water protons in living tissue, the non-invasiveness of the technique and the straightforward spatial encoding of protons by means of additional locally varying magnetic field gradients. During the last two decades, diffusion-weighted imaging (DWI) has added an essential contribution to clinical MR tomography of structural T1- and T2-weighted imaging and to mapping of physiological processes like perfusion and functional activity. DWI uses the self-diffusion (Brownian motion) of water molecules in an inhomogeneous static magnetic field that can be encoded by locally dependent gradient fields. The key feature and importance of MR-DWI results from the fact that the random translatory motion of molecules scans the microscopic tissue structures far beyond the spatial resolution of MR imaging techniques. Furthermore, diffusion as a physical process is independent of magnetic resonance phenomena, but offers similar advantages like high contrast and spatial resolution. The intent of this dissertation covers projects of sequence development and optimization of clinical DWI applications that illustrate the fast evolving development of DWI in medical MR imaging. One focus is set on the combination of DW echo-planar imaging (EPI) and the fluid-attenuated inversion recovery (FLAIR) prepa¬ration. After technical validation, issues of measurement accuracy and signal-to-noise ratio and their implications on estimated contrast parameters like diffusion anisotropy are discussed. A proposed correction term enables immediate acquisition and comparison of standard DW-EPI and FLAIR-DWI in volunteer and patient studies. A preliminary study on patients with astrocytoma reveals the advantages of FLAIR-prepared DW imaging protocols. The second topic of this dissertation explores the extension of the linear diffusion tensor model to high angular resolution DWI (HARDI). The high complexity of white matter

In patients with multiple sclerosis (MS) non-communicating syringomyelia (NCS) has been described as an incidental finding in case studies and small case series. NCS in MS patients commonly leads to uncertainty particularly as the clinical picture of NCS is variable and surgical therapy may be considered. Up to date little is known about the prevalence and clinical importance of NCS in MS. We report the imaging and clinical characteristics of NCS formations in nine MS patients from a 1 year follow-up study in a representative group of 202 MS (4.5%) patients. Brain and spinal cord MRI was performed as part of a genetic study. NCS did commonly extend the central canal and the cord was slightly distended at the level of the syrinx. The cord and syrinx showed no tendency to change in size or shape over 1 year. Despite thorough search into the clinical history and current clinical status no definite but only minimal indications of symptoms potentially related to the NCS were found. We confirm that NCS may occur in MS patients with spinal cord pathology. It can be a subtle finding without clinical correlates. Syrinx formations are more likely to be a consequence of MS cord pathology than a coincidental finding