Fetal brain magnetic resonance imaging (MRI) provides insights into the architecture of the human brain. Recently, an increasing interest has been posed on transient brain structures, such as the ganglionic eminence (GE), to better understand potential derailments or anomalies in neurodevelopment. In this work, we define a spatio-temporal atlas of the GE from 19 to 36 gestational weeks (GW) in a 0.5-mm isotropic resolution. We extended the T2-weighted developing Human Connectome Project atlas with 19 and 20 GW and generated GE label maps spanning 19-36 GW. The GE label maps were generated via an averaging ensemble strategy of the segmentations performed by three expert neuroradiologists. The segmentations conducted by the experts achieved 0.91 ± 0.06 Dice similarity coefficient throughout the whole range of GW, indicating a strong agreement in this task. The GE reached its maximum volume expansion at around 21 GW, followed by a pronounced reduction throughout pregnancy (R2 = 0.98, ranged 40‒500 mm3), highlighting an inverse relationship to the whole brain volume and cortical gray matter. This is accompanied by an increased number of small and fragmented components, correlating with known dynamics of GE migration toward target structures. The proposed spatio-temporal GE MRI atlas supports the monitoring during pregnancy of this fascinating brain structure. It may aid in better understanding prodromic signs of potential future clinical conditions attributable to GE alterations. Moreover, it could be used as a repository of knowledge to develop innovative atlas-based deep learning models for biometric, volumetric, and shape analysis. The spatio-temporal fetal MRI atlas of the GE allows researchers to study its evolution and potential future clinical conditions attributable to GE alterations in pregnancy. The GE reached its maximum volume expansion around 21 GW, followed by a pronounced reduction throughout the pregnancy. The development of GE is a resource for monitoring pregnancy. We propose a spatio-temporal GE MRI atlas from 19 to 36 weeks of gestation. The GE reached its maximum expansion at around 21 weeks of gestation, followed by a progressive decline throughout pregnancy.

We evaluated a microwave imaging prototype to estimate bone-marrow edema in ex vivo bovine samples, using computed tomography (CT) and injected volumes as references. Seven samples comprising distal/proximal halves of two femurs and two humeri were imaged before/after injection of iodine-enhanced gel into drilled holes (Ø 2 mm, depth 8 mm; 25.13 μL/hole). A 4 × 4 open-ended coaxial probe (OECP) array was used over 16 positions with four 90° rotations, simulating a 16 × 16 array of 256 antennas. Raw S11 scattering parameters across a 2.25-3.00 GHz sweep were calibrated and inverted using the Stuchly model with target permittivity at 2.5 GHz. Each OECP antenna was characterized to define its effective penetration depth and equivalent lateral footprint. Upscaled permittivity maps were then generated via nonlinear interpolation, while fluid estimation was computed leveraging a refractive index mixing model. Microwave estimated volumes were repeatable across the four rotations (mean coefficient of variation (CV) 2.8%; mean intraclass correlation coefficient (ICC) 0.999). Microwave estimates tracked the injected volumes with significant correlation (r = 0.81, p = 0.027) and moderate-to-good concordance (Lin's concordance correlation coefficient [CCC] 0.70), with a small bias (+12.4%; 95% limits of agreement (LoA) -17.4% to +42.2%; mean absolute percent error 12.4%). Agreement with CT-derived volumes was weaker (r = 0.58, p = 0.169; CCC 0.23) with a larger bias (+40.9%; 95% LoA -10.3% to +92.2%). Microwave imaging allowed for detecting and quantifying small, μL scale, fluid inclusions within ex vivo long-bone halves, using OECP-based arrays. Non-ionizing microwave imaging may represent a feasible complementary approach for imaging and quantitative assessment of fluid-related distributions in musculoskeletal tissues. A 16-element OECP microwave imaging array allows spatially resolved estimation of fluid-simulated inclusions in ex vivo bovine long bones. Experimental characterization of OECP antenna sensitivity enables resolution upscaling of dielectric permittivity maps. Fluid volumes derived from microwave permittivity maps show significant correlation with injected reference volumes (r = 0.81).

To determine optimal reconstruction parameters for dental implant imaging using photon-counting detector CT (PCD-CT), including ultra-high-resolution (UHR) images, virtual monoenergetic images (VMI), and iterative metal artifact reduction (iMAR). In this ex vivo study, six pig mandibles were prepared with two titanium-based implants and imaged on a PCD-CT. Scans were reconstructed as UHR images and VMI from 70-190 keV at 10 keV increments with and without iMAR. Two independent readers qualitatively evaluated image quality and artifact severity using five-point visual rating scales (5 = excellent, no or minimal artifacts; and 1 = very poor, non-diagnostic, severe artifacts). Two readers quantified artifact severity, defined as the standard deviations in attenuation in regions of interest adjacent to the implants. UHR images without iMAR yielded high image quality (median 5 for both readers) with minor artifact severity (median 4 for both), whereas iMAR reduced image quality (median 3 for both). VMI without iMAR showed decreasing artifacts at higher energy levels. VMI at 120-130 keV achieved optimal image quality (median 5 for both readers at 120 keV, and medians 4 and 5 at 130 keV) with minimal artifacts (median 5 for both), whereas iMAR reduced quality. Quantitative artifact burden decreased with higher energy levels (from 200 HU at 70 keV to 113 HU at 190 keV), and no improvement was observed using iMAR. PCD-CT effectively reduces metal-induced artifacts in dental implant imaging, with UHR images and VMI at 120-130 keV providing optimal image quality, while reconstructions with iMAR offered no further benefit. PCD-CT provides excellent dental implant visualization while minimizing the impact of metal artifacts. In this ex vivo study, ultra-high-resolution images and virtual monoenergetic images at 120-130 keV from PCD-CT effectively reduce metal artifacts from dental implants. Effective artifact reduction offers excellent visualization of the bone-implant interface. Iterative metal artifact reduction (iMAR) did not provide additional benefit for visualization of the bone-implant interface.

Recently, deep learning (DL)-based reconstruction methods have been introduced into clinical magnetic resonance imaging (MRI) systems to enhance image quality and reduce acquisition time. However, their effects on apparent diffusion coefficient (ADC) maps remain unclear. We investigated whether DL-based image reconstruction influences ADC quantification and histogram-based ADC metrics using a calibrated diffusion-weighted imaging (DWI) phantom. A phantom containing vials with known ADC values was scanned on a 3-T system using full (fFOV) and reduced (rFOV) field-of-view DWI sequences. Each acquisition was performed using conventional (DL-OFF) and three DL-based strength levels (low, medium, high). Median ADC values were analyzed for repeatability (coefficient of variation (CV)) and accuracy. Histogram changes and first-order radiomic features were assessed using the Wasserstein distance, Friedman, and Wilcoxon tests. ADC estimates showed high repeatability (CV 0.1-1.2%) and good accuracy (deviation -2 to 7%) across all DL levels and sequences. DL reconstruction progressively reduced histogram dispersion, particularly in high-ADC vials. Wasserstein distances increased with DL strength, confirming a progressive effect on ADC value distributions, while median ADC values remained unchanged. Entropy and interquartile range decreased significantly (p < 0.001), whereas kurtosis and skewness increased, with differences showing less stable and sequence-dependent statistical significance. DL-based reconstruction maintained accurate and repeatable ADC quantification while reducing the dispersion of ADC values. The effect was more evident for high-ADC regions and the rFOV sequence, resulting in narrower distributions of ADC values. Further investigations comparing different DL-based solutions are warranted to assess the generalizability of these findings in clinical settings. Over the past decade, ADC histogram analysis has proven valuable for quantifying tumor heterogeneity, differentiating tumor grade, and evaluating early treatment response. Deep learning reconstruction narrows ADC distributions and reduces dispersion, supporting its potential in oncologic DWI, while highlighting the need for patient-based validation studies. DL reconstruction preserved ADC accuracy in both full FOV and reduced FOV DWI. ADC repeatability remained high across DL levels for both DWI sequences. Histogram dispersion progressively reduced across DL levels, particularly in high-ADC vials. Entropy and interquartile ranges decreased progressively with increasing DL strength.

To evaluate the impact of constraining proprietary and open large language models (LLMs) to structured outputs in processing radiology request forms (RRF). We evaluated five LLMs-two proprietary (GPT-5-Thinking, Gemini 2.5 Pro) and three open (Qwen3-235B-A22B-Thinking, gpt-oss-120b, medgemma-27b-it)-on 100 RRFs (50 computed tomography, 50 magnetic resonance imaging). Each model processed cases with and without constraints to structured outputs. Endpoints included accuracy for modality, anatomical region, contrast phase, urgency, "all correct" (all four categories correct), and "indication improved" (clarity of rewritten text). Outputs were evaluated against a reference standard defined by board-certified radiologists and compared with two radiology residents (first-year and third-year). Accuracies with 95% confidence intervals were calculated. Constraining to structured outputs had model-dependent effects: it improved Gemini 2.5 Pro (all correct: from 53.0% [43.3-62.5] to 66.0% [56.3-74.5]) but reduced GPT-5-Thinking accuracy (from 76.0% [66.8-83.3] to 53.0% [43.3-62.5]), with minimal influence on open models. Both proprietary LLMs outperformed the best open models (up to 41.0% [31.9-50.8]). All LLMs exceeded the unassisted first-year residents' performance (19.0% [12.5-27.8]). LLM assistance improved first-year residents' accuracy to 65.0% [55.3-73.6], approaching the third-year residents' performance (80.0% [71.1-86.7]), who performed comparably to the best LLMs. Across models, performance was highest for modality and anatomical region, and lowest for urgency. Indication reformulation was judged clearer in > 90% of cases across all models without hallucinations. Constraining to structured outputs exerted model-specific effects. Proprietary LLMs achieved the highest accuracy in RRF-based protocol selection and improved first-year resident performance to an experienced-resident level. LLMs may serve as valuable decision-support tools for radiology workflow. Constraining LLMs to structured outputs produced divergent, model-specific effects in radiology protocol selection-improving Gemini 2.5 Pro, reducing GPT-5-Thinking, and minimally affecting open models-highlighting the need for model-specific prompting strategies before adopting LLMs in radiology decision support. Structured output constraints affect LLM performance differently. Gemini 2.5 Pro benefits from structured prompting, while GPT-5-Thinking declines. Open-weight models show minimal impact from output constraining. Proprietary models outperform open models in radiology protocol selection.

Liver injury includes inflammation and resolution stages with different macrophage populations. Hepatic macrophages can be imaged with ultrasmall paramagnetic iron oxide (USPIO) nanoparticles-enhanced magnetic resonance imaging (MRI). We aimed to assess if inflammation and resolution stages of liver injury could be differentiated with USPIO-enhanced MRI in mice. Three groups of C57BL/6JRj mice (control, inflammation, resolution; n = 10 for each group) were imaged. Liver fibrosis was induced by intraperitoneal carbon tetrachloride injections for 6 weeks. Multigradient-echo MRI was performed before, 24, and 48 h after intravenous injection of fluorescent USPIO. Contrast uptake was quantified with measurements. Macrophages were immunostained with F4/80, and USPIO fluorescence was localized and quantified with confocal microscopy. Liver iron content was measured with inductively coupled mass spectrometry (ICP-MS). was assessed with Mann-Whitney tests. Kruskal-Wallis and Dunn tests compared fibrosis, fluorescence, and ICP-MS iron concentration. Spearman tests were used for correlation analysis. was significantly higher in the inflammation group (158 ±85%) compared to the control (58 ± 36%, p = 0.020) and resolution (71 ± 36%, p = 0.048) groups. Confocal microscopy showed a high macrophage number and USPIO uptake in the inflammation group. correlated with macrophages number (r = 0.67, p = 0.0001), USPIO fluorescence intensity (r = 0.58, p = 0.0011), and iron concentration at ICP-MS (r = 0.39, p = 0.028). Our results suggest that the inflammatory and resolution stages of hepatic injury can be assessed with USPIO-enhanced MRI. Our study demonstrates that USPIO-enhanced MRI can be used to monitor the inflammatory and resolution phases of hepatic injury in mice. If future studies confirm these findings, this imaging method might be valuable for tracking hepatic inflammation dynamics. Liver was highest during the inflammatory stage and partially reversed during the resolution stage, reflecting macrophage dynamics. Immunofluorescence showed increased macrophage number and uptake during inflammation, decreasing during resolution. Iron concentration significantly correlated with , macrophage number, and total fluorescence intensity.

Fine-needle aspiration biopsy (FNAB) and core-needle biopsy (CNB) are common methods to evaluate the pathology of an abnormal tissue mass. However, both methods have their limitations, e.g., FNAB samples often remain inadequate, and CNB is more invasive. A novel device, ultrasound-enhanced fine-needle aspiration biopsy (USeFNAB), can collect samples with greater cellular content quantitatively without compromising tissue quality. The purpose of this prospective study was to evaluate USeFNAB for the first time in humans in vivo. This is a pilot study without any deviation from the contemporary management of the salivary gland tumor. Ten adult patients with a solid benign parotid gland tumor diagnosed by FNAB or CNB. Before parotidectomy, the tumors were sampled under ultrasound guidance with three different needle sampling techniques: USeFNAB, FNAB, and CNB. The influence of USeFNAB on the quantity and quality of the samples was investigated and compared with FNAB and CNB. The quality of the cytological slides and histological tissue samples was similar in all samples obtained by USeFNAB, FNAB, and CNB. With USeFNAB, the mass increased on average by 1.6 and 3.4 times and the histological sample area by 1.7 and 3.4 times, as compared to FNAB and CNB, respectively. The results of this study demonstrated that USeFNAB seems to be a feasible and safe biopsy technique under in vivo conditions. In solid benign parotid gland tumors, USeFNAB increases sample yield as compared to conventional needle sampling methods, without affecting sample quality. USeFNAB could improve the diagnostic accuracy of needle biopsies and facilitate ancillary techniques such as molecular diagnostic studies. Institutional permissions for the study were granted. The USeFNAB protocol went through risk analysis and mitigation for approval by the Finnish Medicine Agency (FIMEA) as an investigational device for research purposes (FIMEA/2023001788). USeFNAB showed an improved sample yield in solid benign parotid gland tumors compared to FNAB and CNB, without affecting sample quality. USeFNAB seems to be a feasible and safe biopsy technique. USeFNAB is a promising tool for improving the diagnostic accuracy of needle biopsies.

We investigated the extent of cellular, transcriptional, and translational activation throughout the liver following radiofrequency ablation (RFA). RFA of the healthy liver was performed in two 8-10-week-old male C57/Bl6 mice, no/sham procedure in one. One and 7 days after, single-cell RNA sequencing (scRNAseq) was performed on distant, untreated liver to examine > 6,000 genes from normalized datasets of > 6,000 cells/sample, enabling identification of ten major cell populations. We defined cell-to-cell interactions by CellphoneDB and identified active pathways via STRING-db analysis with Markov clustering. Twelve distant liver lobe samples were homogenized on day 3 or day 6 after RFA/sham procedure for SomaLogic proteomic analysis (> 1,300 genes), subsequent STRING-db analysis, and assessment of cellular origin (PanglaoDB-2021). CellphoneDB identified crosstalk among all ten populations with 4,658 and 4,218 receptor/ligand pairs, identified on day 1 and day 7 post-RFA, respectively. On day 1, 360 differentially expressed genes were identified; on day 7, 430. Activated genes distributed into 16 clusters, including 66 chemokines/cytokines, including Ccl2 and Ccl7; 57 immunomodulators, including Il6, Ctla4 and Pdcd1; and 54 growth factors, including Vegf, Hgf, Pdgf, and Fgf. Angiogenesis pathway genes were observed in endothelial cells and hepatocytes. Pdcd1 and Ctla4 were notably increased transiently in T cells. Proteomic analysis included 228/443 genes (51%) identified by scRNAseq; 73/228 proteins (32%) demonstrated 25% elevation over controls. Overall, 427 proteins were elevated, with 9/10 cell populations contributing to increased protein expression (odds ratio 4.9‒7.0). RFA diffusely activates cellular processes remotely from the ablation zone on both transcriptional and translational levels, altering tumorigenic and immunologic pathways simultaneously. This study offers insights into liver tissue biology after RFA and provides a comprehensive picture of the molecular mechanisms put into motion by this procedure. A better understanding of these processes could provide a potential basis to develop specific biomarkers and effective adjuvant therapies following local tumor ablation. RFA activates a multiplicity of hepatic cellular processes remotely from the ablation zone on a transcriptional and translational level. Single-cell RNA sequencing provides insights into widespread cellular origins of activated pro-immunogenic, pro-tumorigenic, and other pathways detected post-ablation. Consideration of the nature of this response may help achieve the clinical goals of adjuvant therapies and predictive biomarkers.

To evaluate whether 7-T susceptibility-weighted imaging (SWI) can predict glioma's histological grade, Ki-67 labeling index (LI), isocitrate dehydrogenase 1 (IDH1) mutation, 1p/19q co-deletion, telomerase reverse transcriptase (TERT) promoter mutation, and O-6-methylguanine-DNA-methyltransferase (MGMT) promoter methylation status of gliomas. We retrospectively analyzed 7-T SWI in 60 patients with glioma. The Mann-Whitney U test compared the intratumoral susceptibility signals (ITSS) grade across molecular markers, with ITSS defined as fine linear or dot-like low signal areas on SWI. Predictive efficacy was assessed using receiver operating characteristic (ROC) analysis and multivariate logistic regression models. Path analysis evaluated the relationships between ITSS grade and molecular markers. Gliomas with high ITSS grade showed higher histological grade, Ki-67 LI, and TERT mutation rates compared to those with low ITSS grade, mostly being wild-type gliomas. ITSS grade predicted the histological grade, Ki-67 LI, and TERT status (area under the ROC curve = 0.769‒0.817). Multivariate logistic regression analysis identified Ki-67 LI and TERT status as independent predictors of high ITSS grade. Path analysis indicated direct effects of Ki-67 LI and TERT mutation on ITSS grade, and an indirect effect of IDH1 mutation on ITSS grade mediated through Ki-67 LI. 7-T SWI-derived ITSS grade predicts histologic grade, Ki-67 LI, and TERT promoter mutation status in gliomas. Ki-67 LI and TERT mutation exert relatively independent effects on ITSS grade and allow reverse inference of their status from SWI, whereas IDH1 mutation influences ITSS grade indirectly via Ki-67 LI. This study establishes a connection between preoperative imaging and molecular glioma pathology via 7-T SWI. It helps to reveal the in vivo characteristics of pathology and promotes collaboration among radiologists, pathologists, and clinicians, which a great clinical potential. A 7-T susceptibility-weighted MRI-based intratumoral susceptibility signal (ITSS) grading system enables precise detection of glioma microbleeds and neovascularization. 7-T susceptibility-weighted MRI-derived ITSS grade noninvasively predicts histologic grade, Ki-67 labeling index, and telomerase reverse transcriptase (TERT) promoter mutation status in gliomas. Path analysis suggested that molecular markers relate to ITSS grade through distinct pathways, with Ki-67 and TERT exerting direct effects and isocitrate dehydrogenase 1 influencing ITSS grade indirectly.

Photon-counting detector CT (PCD-CT) enables spectral imaging with material separation. Accurate iodine and iron quantification remains challenging due to inevitable low- and high-energy base material CT number mismatches and dual-energy ratio (DER) variability. This study develops and validates a correction method addressing these issues to improve iodine and iron quantification in PCD-CT. A spectral CT abdomen phantom containing rods with known iodine (0.5-15.0 mg/mL) and iron (2.0-25.0 mg/mL) concentrations in water- and liver-equivalent material was scanned on a clinical PCD-CT under varying tube voltages, dose levels, and with/without a fat ring. High- and low-energy CT numbers of base materials and DER values were inputs for the correction method. Material concentrations calculated with and without correction were validated against known phantom values. The correction method significantly reduced quantification errors. Iodine errors fell below 5% for concentrations ≥ 2 mg/mL and iron errors below 15% for concentrations ≥ 5 mg/mL. Without correction, errors reached up to 83% (iodine) and 85% (iron) at low concentrations, reduced to 23% and 47%, respectively, after correction. The proposed correction method improves accuracy in spectral material decomposition for PCD-CT, supporting its potential for better clinical assessment of lesion contrast enhancement, therapy response and hepatic burden evaluation. This technical note introduces a phantom-based correction method for photon-counting detector CT that improves iodine and iron quantification by addressing base material Hounsfield Unit (HU) mismatches and dual-energy ratio variability. The method reduces quantification errors and offers a practical calibration procedure, supporting the potential for clinically reliable iodine and iron quantification. PCD-CT correction method reduces concentration errors across varying scan protocols and configurations. Implementation guide supports adaptation to other scanners. Accurate iodine and iron quantification supports diagnosis and treatment assessment.