Steven Hubert Baete

Diffusion Magnetic Resonance Imaging (dMRI) is a noninvasive modality that enables the study of brain tissue microstructure and the reconstruction of neural pathways. To achieve this, most reconstruction methods rely on inverse modeling techniques that fit the measured signal to narrow parametric spaces, each designed to recover either fiber orientations or microstructural properties. Obtaining a comprehensive characterization of tissue microstructure and architecture therefore often requires combining multiple models, which is computationally demanding and potentially inconsistent due to model-specific assumptions and acquisition settings. This work introduces FORCE, a forward modeling paradigm that reframes how diffusion data is analyzed. Instead of inverting the measured signal, FORCE simulates a large set of biologically plausible intra-voxel fiber configurations and tissue compositions. It then identifies the best-matching simulations for each voxel by operating directly in the signal space, retaining a local posterior over plausible configurations. This unified framework performs three tasks in a single process: resolving low-angle fiber crossings, generating a large suite of microstructural maps, and producing partial volume maps, while also yielding a per-voxel measure of estimation uncertainty. The proposed approach demonstrates robust performance across synthetic and real datasets from both human and mouse brains, encompassing multiple resolutions and acquisition types, and requires no specialized acquisition or additional scan time.

Purpose: In diffusion MRI, vasogenic edema manifests as a major fraction of isotropic water that dilutes the anisotropic intra‐axonal portion of the signal. Many tractography algorithms mistake vasogenic edema for the white matter boundary and terminate tracking to prevent producing spurious streamlines. As a result, visual representations of fascicles traversing edema are often compromised, limiting the clinical utility of tractography. Methods: We address this hurdle with ODF‐Fingerprinting (ODF‐FP)—a dictionary‐based fiber reconstruction algorithm that accommodates variability of neural tissue. By adding a regularization term to the ODF‐FP matching formula, we counterbalance the drop of diffusion anisotropy in edematous regions to improve white matter fiber identification. In 19 glioma cases with significant peritumoral vasogenic edema, we quantify the volume of the reconstructed white matter tracts immersed in edema, then we use the cortical regions activated during task‐based functional MRI as validation for tractography. To assess the potential for clinical translation, we additionally test the performance of ODF‐FP on subsampled single‐shell diffusion‐weighted images, which contemporary clinical scanners can acquire within a few minutes. Results: Our approach produces high volumes of streamlines traversing vasogenic edema and reaches high overlap with the cortical regions activated at task‐based fMRI, significantly outperforming common fiber reconstruction methods in the clinically feasible data set. Conclusion: ODF‐FP proves effective on research and clinical quality dMRI, which offers an opportunity for application in neurosurgery.

Optimal treatment of glioma has been a subject of debate over the last few decades, since maximal resection can improve survival, whereas preservation of functional peritumoral brain tissue minimizes the risk of postoperative neurological deficits. Our preliminary study uses tractography and neural tissue microstructure modeling based on diffusion MRI to quantify progressive axonal degeneration in proximity to frontotemporal glioma. For this, we sample major white matter pathways that traverse peritumoral penumbra at two time points. The results show a pattern of decreased intra-axonal water fraction beyond anatomical MRI abnormalities, which may indicate a tumor invasion of normal-appearing white matter that potentially advocates supratotal resection.

Diffusion Magnetic Resonance Imaging (dMRI) is a noninvasive modality that enables the study of brain tissue microstructure and the reconstruction of neural pathways. To achieve this, most reconstruction methods rely on inverse modeling techniques, which are often ill-posed and struggle to resolve shallow fiber crossings. Moreover, existing methods typically focus either on estimating fiber orientations or on deriving microstructural maps. As a result, obtaining a comprehensive characterization of tissue microstructure and architecture often requires combining multiple models, which is computationally demanding, potentially inconsistent due to model-specific assumptions and acquisition settings. This work introduces FORCE, a forward modeling paradigm that reframes how diffusion data is analyzed. Instead of inverting the measured signal, FORCE simulates a large set of biologically plausible intra-voxel fiber configurations and tissue compositions. It then identifies the best-matching simulation for each voxel by operating directly in the signal space. This unified framework simultaneously resolves low-angle fiber crossings, producing a large suite of microstructural maps and complete tissue segmentation in a single process. The proposed approach demonstrates robust performance across synthetic and real datasets from both human and mouse brains, encompassing multiple resolutions and acquisition types.

In diffusion MRI, vasogenic edema manifests as a major fraction of isotropic water that dilutes the anisotropic intraaxonal portion of the signal. Many tractography algorithms mistake vasogenic edema for the white matter boundary and terminate tracking to prevent producing spurious streamlines. As a result, visual representations of fascicles traversing edema are often compromised, limiting the clinical utility of tractography in neurosurgery. We address this hurdle with ODF-Fingerprinting (ODF-FP) — a dictionary-based fiber reconstruction algorithm that accommodates variability of neural tissue. By adding a regularization term to the ODF-FP matching formula, we counterbalance the drop of diffusion anisotropy in edematous regions to improve white matter fiber identification. We test our approach in 19 glioma cases with significant peritumoral vasogenic edema. For each case, we quantify the volume of the reconstructed white matter tracts immersed in edema, then we use the cortical regions activated during task-based functional MRI as validation for tractography. To assess the potential for clinical translation, we additionally test the performance of our reconstruction method on subsampled single-shell diffusion-weighted images, which contemporary clinical scanners can acquire within a few minutes.

Peritumoral vasogenic edema of the brain is a major confounding factor for diffusion MRI tractography. Excessive fluids accumulated in edematous white matter decrease anisotropy of water self-diffusion which affects tracking algorithms. We address this hurdle with ODF-Fingerprinting (ODF-FP) — a dictionary-based fiber reconstruction algorithm that accommodates variability of neural tissue. By adding a regularization term to the ODF-FP matching formula, we boost diffusion anisotropy to improve white matter fiber identification in edematous regions.

Magnetic resonance imaging (MRI) has experienced remarkable advancements in the integration of artificial intelligence (AI) for image acquisition and reconstruction. The availability of raw k-space data is crucial for training AI models in such tasks, but public MRI datasets are mostly restricted to DICOM images only. To address this limitation, the fastMRI initiative released brain and knee k-space datasets, which have since seen vigorous use. In May 2023, fastMRI was expanded to include biparametric (T2- and diffusion-weighted) prostate MRI data from a clinical population. Biparametric MRI plays a vital role in the diagnosis and management of prostate cancer. Advances in imaging methods, such as reconstructing under-sampled data from accelerated acquisitions, can improve cost-effectiveness and accessibility of prostate MRI. Raw k-space data, reconstructed images and slice, volume and exam level annotations for likelihood of prostate cancer are provided in this dataset for 47468 slices corresponding to 1560 volumes from 312 patients. This dataset facilitates AI and algorithm development for prostate image reconstruction, with the ultimate goal of enhancing prostate cancer diagnosis.

Purpose: The accuracy of diffusion MRI tractography reconstruction decreases in the white matter regions with crossing fibers. The optic pathways in rodents provide a challenging structure to test new diffusion tractography approaches because of the small crossing volume within the optic chiasm and the unbalanced 9:1 proportion between the contra‐ and ipsilateral neural projections from the retina to the lateral geniculate nucleus, respectively.Methods: Common approaches based on Orientation Distribution Function (ODF) peak finding or statistical inference were compared qualitatively and quantitatively to ODF Fingerprinting (ODF‐FP) for reconstruction of crossing fibers within the optic chiasm using in vivo diffusion MRI ( healthy C57BL/6 mice). Manganese‐Enhanced MRI (MEMRI) was obtained after intravitreal injection of manganese chloride and used as a reference standard for the optic pathway anatomy.Results: ODF‐FP outperformed by over 100% all the tested methods in terms of the ratios between the contra‐ and ipsilateral segments of the reconstructed optic pathways as well as the spatial overlap between tractography and MEMRI.Conclusion: In this challenging model system, ODF‐Fingerprinting reduced uncertainty of diffusion tractography for complex structural formations of fiber bundles.