Publications

Objective To evaluate how data acquisition and post-processing choices influence ultrashort echo time (UTE)–based characterization of the short-T2* compositional structure of the Achilles tendon. Methods Ten healthy volunteers (33.3 ± 6.9 years) underwent Achilles tendon MRI using a multi-echo, time-interleaved UTE sequence. T2* data were analyzed with and without fat suppression using two models: (i) a bi-component decay with two water pools and (ii) a tri-component decay including two water components and one fat component. The effect of magnitude versus complex fitting on compositional estimates was also assessed. Three Achilles tendon regions were analyzed: insertion (INS), mid-portion (MID), and muscle–tendon junction (MTJ). Results Estimated short vs. long T2* components differed significantly depending on the use of fat suppression (p < 0.001 for all tendon regions) and fitting method (magnitude vs. complex) (pINS = 0.002, pMID < 0.001, pMTJ < 0.001). T2* time constants were comparatively stable, particularly in the MID part (p > 0.05). The fitting method affected short T2* component in the MTJ (pMTJ = 0.032) and long T2* component in the INS and MTJ (pINS = 0.025, pMTJ < 0.017). Discussion Non-fat-suppressed vs. fat-suppressed (FS) strategy and fitting approach substantially affect bi-component T2* quantification of the Achilles tendon, limiting comparability of UTE-based metrics. Tri-component modeling provides a more realistic description of tendon signal and may improve sensitivity to pathological changes. Further studies in patients are needed.

Purpose Accurate arterial input functions (AIFs) are essential for quantitative dynamic contrast-enhanced (DCE) MRI, yet direct measurement is challenging and population-averaged AIFs neglect patient-specific variability. Blind deconvolution provides an alternative by estimating patient-specific AIFs directly from tissue data, up to a scale factor. This study compared blindly estimated AIFs with carefully measured aortic AIFs in breast DCE-MRI. Theory and Methods Data from 25 patients with breast cancer were analyzed, each with a carefully measured AIF. Blind AIF estimates were obtained using model-constrained deconvolution in the Perflab toolkit with both Tofts–Kety (TK) and two-compartment exchange (2CXM) tissue models. To help isolate AIF shape blind AIFs were scaled using cardiac output. These blind AIFs were compared with measured AIFs using scale-invariant and scale-dependent metrics which assess the similarity of the AIF shape. Results Blind estimates were obtained for all 25 patients with both tissue models. Compared with measured AIFs, blind estimates derived using the 2CXM showed stronger agreement across all metrics than those derived using the TK model. However, the weak correlation in scale-dependent metrics suggests limitations of cardiac output scaling. Conclusion Blind AIF estimation using the 2CXM provides more reliable recovery of AIF shape and dispersion than the TK model in breast DCE-MRI. While blind deconvolution shows promise for estimating local patient-specific AIFs other scaling strategies may need to be employed in practice.

Object This study exploits newly developed dynamic indirect 1H-[2H]-FID-MRSI at 9.4 T, combined with a dedicated metabolic model, to enable regional and quantitative characterization of glucose oxidative metabolism flux in the rat brain with minimal metabolic assumptions, by measuring both 2H-labelled Glx turnover and pool size along a controlled 2H-Glc infusion protocol. Materials and Methods Seven rats underwent dynamic 2D 1H-FID-MRSI during a 2-h infusion of [6,6’-2H₂] glucose. Consecutive 13-min acquisitions quantified Glx-C4 1H-signal decay, converted to 2H-Glx concentrations using baseline metabolite pool sizes. A four-pool kinetic model including 2H-label loss was fitted to regional turnover curves to estimate oxidative flux (Vgt) and pyruvate dilution (Kdil). Model performance and parameter robustness were assessed with Monte-Carlo simulations. Results In vivo 2H-Glx turnover showed a saturated exponential rise (~ 60 min), with a labelling plateau higher in striatum (1.85 μmol/g) than hippocampus (1.55 μmol/g). Metabolic modelling provided region-specific oxidative fluxes: Vgt= 0.53 ± 0.15 μmol/g/min (hippocampus) and Vgt= 0.81 ± 0.12 μmol/g/min (striatum), with consistent Kdil across regions. Simulations confirmed a good model robustness in retrieving Vgt over a large range of experimental conditions. Discussion This work shows the potential of indirect dynamic 1H-[2H]-FID-MRSI for quantitative metabolic flux mapping of cerebral glucose oxidative metabolism.

The unique capability of deuterium metabolic imaging (DMI) to detect downstream metabolic products and trace substrates’ transport within tissues using conventional magnetic resonance (MR) scanners can, in theory, be employed with routine positron emission tomography (PET)/MR equipment. Our technical proof-of-concept study proposes a protocol for the simultaneous acquisition of DMI and [¹⁸F]FDG-PET data to enable dual assessment of hepatic glucose metabolism. A protocol that integrates high-dose glucose administration, required for DMI, with [¹⁸F]FDG-PET imaging was applied in a spectroscopy-validated rodent model of metabolic dysfunction-associated fatty liver disease (MAFLD). We acquired and quantified high-quality DMI and PET data of the liver that could provide a distinction between healthy and MAFLD cohorts in the future.

IVIM tensor imaging is presented to quantify the directional microvascular architecture in vivo. Perfusion tensors from synthetic data, a phantom with aligned fibers, and preclinical rat measurements reveal a microvascular orientation. Diffusion tensors capture tissue-dependent microstructural heterogeneity. The method distinguishes tissues with chaotic versus predominantly oriented microvessels and identifies orientation patterns, providing a novel biomarker for non-invasive in vivo characterization of microvascular heterogeneity. These findings support its potential for tumor staging, treatment monitoring, and early detection of brain disorders.

Aims This study employed diffusion kurtosis imaging (DKI) and proton magnetic resonance spectroscopy (1H-MRS) on an MPTP-induced mouse model of Parkinson's disease (PD) to examine microstructural changes linked to neuroinflammation and neurodegeneration. Methods MPTP (20 mg/kg, i.p.) was given for 4 days, and behavioral assessment, MRI imaging, and immunohistochemistry were performed at 24 h and 72 h after last MPTP treatment. Results At 24 h, DKI showed higher diffusivity metrics in the hippocampus and thalamus, while 1H-MRS identified reduced Glu/tCr and Glx/tCr ratios in the striatum of MPTP-treated mice compared to saline-treated mice. Behavioral tests at 72 h revealed motor impairment and DKI showed increased diffusivity in the somatosensory cortex, thalamus, and striatum in MPTP-treated mice. Notably, at 72 h, the hippocampus showed partial recovery in diffusivity, suggesting adaptive changes or partial restoration. Higher diffusivity was observed in the cortex, striatum, and thalamus in MPTP-treated mice. Furthermore, 1H-MRS detected a higher Tau/tCr in the striatum, while in the hippocampus, lower Gln/tCr and NAA/tCr and higher Cho/NAA were observed at 72 h in MPTP-treated mice, indicating persistent neuronal death and membrane deterioration. Immunofluorescence staining at 72 h confirmed these findings, showing a decrease in NeuN+ neurons and an increase in GFAP+ glial cells in the striatum and hippocampus, indicating neurodegeneration and gliosis. Additionally, MPTP caused a loss of dopaminergic neurons in the substantia nigra and striatum, which likely explains the higher diffusivity shown by DKI. Conclusion These findings demonstrate DKI and 1H-MRS are sensitive, non-invasive modalities for detecting and monitoring neurodegenerative microstructural and neurochemical changes, enhancing the understanding of PD-related pathology and progression.

Accepted at IEEE ISBI 2026. This is the author's accepted manuscript version. Objective and reproducible evaluation of magnetic resonance imaging (MRI) quality remains challenging in the absence of reliable reference data. We introduce MR-IQ, a novel open-source, reference-free metric for quantitative MRI quality assessment. MR-IQ combines three normalized sub-metrics: the normalised signal-to-noise ratio (nSNR), sharpness index (SI), and artefact-free index (AFI). These three sub-metrics are integrated into a single composite score via a harmonic mean formulation. Experiments on preclinical MRI datasets with simulated noise, blur, and motion artefacts demonstrate that MR-IQ provides consistent and interpretable quality scores aligned with human perception, outperforming conventional blind metrics such as BRISQUE. The proposed metric enables quantitative, artefact-sensitive, and reproducible evaluation of MRI data, facilitating its integration into automated quality-control pipelines and AI-based reconstruction workflows.

Neuropsychiatric malignancies frequently manifest at the level of individual cortical layers. The resolutions currently available for medical magnetic resonance imaging (MRI) prevent the study of these pathologies at clinically available field strengths of 3 T. Previous studies have claimed to have overcome these issues by extensions of quantitative MRI. Following this, the feasibility of multiexponential T1 relaxometry was assessed as a basis for in vivo delineation of cortical lamination. Three methods of non-linear least-squares-based multiexponential analysis were examined across key degrees of freedom identified in the literature. The methods employ a wide variety of ways to overcome the common pitfalls of multiexponential analysis, such as regularization, bound constraints, and repeated optimization from multiple starting points. A custom MRI phantom was 3D-printed and filled with various MnCL2 mixtures that represent the spin-lattice relaxation times that commonly occur in neocortical gray and white matter at 3 T. A 96 × 96-voxel image consisting of a single slice was acquired using a FLASH sequence and used to create 10 composite datasets with known distributions of T1 decay constants. The results showed that lowest relative error achieved across multiexponential models was approximately 20%. As achieving even this level of estimation accuracy requires either T1 ratios that rarely occur in the cerebral cortex or knowledge of the number of relaxation components and their expected values to a degree that is seldom feasible, the visualization of cortical layers based on these estimates is unlikely to represent their true distribution. In conclusion, the current methodological approaches do not allow for sufficiently precise estimation of T1 decay constants spanning the range of cortical gray and white matter.

Background: The heterogeneous diagnostic methods for sarcopenic obesity lack standardization across clinical practice. The 2022 ESPEN/EASO consensus established standardized diagnostic criteria emphasizing weight-adjusted muscle mass indices, yet implementation feasibility in bariatric surgery populations remains unexplored. The absence of validated cutoff values for Class III obesity and limited evidence on dual-modality assessment protocols represent critical barriers to clinical implementation. Objective: This pilot study objectively assessed the implementation of ESPEN/EASO consensus criteria using dual-modality DXA and MRI assessment in women undergoing sleeve gastrectomy, with primary focus on protocol completion rates, technical precision, and preliminary evaluation of weight-adjusted diagnostic indices. Methods: Following CONSORT 2010 pilot trial extension guidelines, this feasibility study analyzed eleven women (age 41.3 ± 7.9 years (95 % CI: 36.6–46.0 years), 41.0 ± 5.0 kg/m2 (95 % CI: 38.0–44.0 kg/m2) undergoing laparoscopic sleeve gastrectomy. Primary feasibility outcomes included dual-modality assessment completion rates, MRI segmentation algorithm performance, and protocol adherence. Secondary outcomes encompassed preliminary body composition changes and ASM/weight diagnostic utility evaluation using ESPEN/EASO recommended cutoffs. Results: Dual-modality assessment achieved 100 % completion rates with excellent technical precision (MRI segmentation Dice coefficient: 0.95 ± 0.01 for muscle, 0.95 ± 0.03 for adipose tissue). Protocol adherence was optimal with all participants completing scheduled assessments. Preliminary body composition analysis demonstrated substantial weight reduction of 29.4 ± 25.3 kg (95 % CI: 14.4–44.4 kg) at six months. ASM/weight ratios showed potential diagnostic utility, though wide confidence intervals reflect substantial uncertainty inherent in small sample analysis. Conclusions: Implementation of ESPEN/EASO consensus criteria demonstrates 100 % protocol completion rates and MRI segmentation precision with Dice coefficients of 0.95 ± 0.01 for muscle and 0.95 ± 0.03 for adipose tissue in bariatric surgery populations. This pilot study establishes the methodological foundation and demonstrates feasibility for larger-scale validation studies to develop diagnostic thresholds and optimize sarcopenic obesity assessment protocols in bariatric medicine.

Purpose Quantifying proton density fat fraction (PDFF) in small abdominal organs is challenging due to low T1/T2 contrast and susceptibility artifacts. We develop a hybrid 7-echo CSE-MRI sequence with arbitrary echo spacing, inspired by GRASE-type imaging, aiming for distortion-free PDFF mapping in small animals. The method is designed to be comparable to established conventional methods, with potential for increased robustness. Methods We developed a Fast Spin Echo Asymmetric Bipolar Multi-Gradient Echo (FSE-AbMGE) sequence by integrating a fast spin-echo readout with an asymmetrically placed bipolar multi-echo gradient-echo train. The sequence was implemented at 9.4 T and combined with robust phase unwrapping and water-fat reconstruction algorithms using full fat spectral modeling. Validation was performed using phantoms with known PDFF values (0–22 %) and in vivo experiments on several female mice (n = 2). Reference PDFF values were obtained using single-voxel 1H-MRS. Results The proposed method enabled high-resolution PDFF mapping with minimal chemical shift and susceptibility artifacts. Phantom experiments showed strong agreement with both spectroscopic and ground truth values (R2 > 0.98, p < 0.001). The method was also tested in vivo, demonstrating robust water-fat separation and quantification. Conclusion The FSE-AbMGE sequence is well-suited for accurate abdominal fat quantification in small animals. While additional validation is needed, especially in reproducibility and broader biological settings, the method shows promise for high-field fat quantification and may offer a framework adaptable to lower-field pre-clinical applications.