European Radiology Experimental最新文献

Objective: This study explored the gradient changes in structural similarity based on the cortical structure of patients with lifelong premature ejaculation (LPE) and further analyzed the characteristics of the associations between these changes and clinical phenotypes, gene expression profiles, and neurotransmitter distributions.

Materials and methods: This study employed a novel method, morphological inverse divergence (MIND), to construct structural similarity gradients for 62 LPE patients and 53 healthy controls. Between-group comparisons were performed to examine the abnormalities in gradients among LPE patients. Partial least squares regression analysis explored the relationships between gene expression profiles and gradient changes, as well as neurotransmitter expression associations with these alterations.

Results: We found that both groups showed a classic unimodal-to-cross-modal transition along the principal gradient. In LPE patients, the principal gradient increased in the left visual cortex and right prefrontal regions but decreased in the right cingulate gyrus. The secondary gradient also decreased in the right somatosensory cortex and bilateral visual cortices. Notably, changes in these gradients in the right somatosensory and visual cortices were significantly negatively correlated with clinical phenotypes. Connectome-transcriptome analysis revealed that abnormal gradient patterns were linked to whole-brain gene expression profiles, with enriched genes in pathways related to hormone activity and other functions. Additionally, there was a spatial correlation between the gradients and neurotransmitter densities.

Conclusion: We identified the biological pathways enriched in genes associated with the pathological process of LPE and characterized the distribution patterns of neurotransmitter receptors and transporters, thereby providing critical insights into the neuroimaging and neurobiological underpinnings of LPE.

Relevance statement: The MIND-based brain structural similarity gradient exhibits a pattern of segregation and integration. We analyzed the association between this structural gradient and the clinical phenotype of primary premature ejaculation, offering novel insights into the neuroimaging and neurobiological mechanisms underlying the disorder.

Key points: We employed a novel approach, morphometric inverse divergence, to construct the brain structural similarity gradient. We provide evidence for abnormal structural similarity gradients in patients with LPE. We found an association between abnormal changes in gradients and clinical phenotypes, gene enrichment pathways, as well as neurotransmitter density.

Background: To increase the number of patients who can speak hands-free after total laryngectomy, we aim to assess the viability of the manubrium sterni bone (MSB) for bone implantation.

Materials and methods: This retrospective cross-sectional study was conducted at the Department of Head and Neck Surgery of the Netherlands Cancer Institute in Amsterdam, the Netherlands, after approval by its Institutional Review Board. We included CT scans of 25 males and 24 females aged 53 to 88 years, and measured MSB morphometry and density, soft tissue thickness (STT), and bone surface slope, and classified the MSB. Heat maps were created to identify potential implant locations.

Results: Median MSB height was 50.7 [95% confidence interval 49.4-53.4] mm, and mean thickness ranged from 8.7 [8.3-9.2] to 13.1 [12.7-13.6] mm. Mean width ranged from 53.2 [51.1-55.4] to 57.1 [55.3-59.1] mm. Significantly greater values of thickness and width were found in males. Body height had a significant positive effect on thickness and width measured at the thickest level. Median STT was 14.4 [12.7-17.4] mm, on which BMI had a significant positive effect. The trapezoid was the most common shape. Mean densities were 325 and 60 HU for cortical and cancellous bone, respectively, and cancellous density was significantly higher in males. As such, the MSB is most similar to the posterior maxilla with regard to implantation.

Conclusion: In theory, the MSB is able to support a combination of three to four 6-mm implants, or of one 6-mm to 8-mm and two 10-mm implants, placed in a linear, triangular or diamond-shaped arrangement.

Relevance statement: This study addresses knowledge gaps regarding the possibility of fixation methods of hands-free speaking valves offering greater stability than conventional methods. Radiological assessment of CT scans of the manubrium sterni bone demonstrates its potential as an implantation site, although a thorough preimplantation workup of individual anatomy is imperative.

Key points: This is the first study that analyzes the manubrium sterni bone three-dimensionally and assesses its potential as a bone implantation site. We found that it has low cortical and cancellous bone density, which makes it most similar to the posterior maxilla. The Manubrium sterni bone is suitable for implantation, although a preimplantation assessment of individual anatomy is imperative.

Background: Detection of axillary lymph node (LN) involvement is essential for staging breast cancer and optimizing treatment. This proof-of-concept two-center study explored the feasibility of magnetic resonance imaging (MRI) radiomics-based machine learning models to predict LN involvement and compare their performance with node reporting and data system (Node-RADS).

Materials and methods: We retrospectively included breast cancer patients undergoing preoperative multiparametric MRI and LN dissection (January 2020-September 2024). Stable radiomic features (intraclass correlation coefficient ≥ 0.75) were extracted from contrast-enhanced, subtracted, and T2-weighted sequences. Five machine learning models were trained for binary LN involvement classification, using histopathology as a reference standard. The best-performing model was externally validated on an independent cohort. Performance metrics included sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUROC). Node-RADS (scores > 2 indicating LN involvement) was used for comparison in the external dataset.

Results: Of 93 cases, 40 (43%) were LN involvement-positive; 17 stable features were selected for model development. The best-performing model achieved 81% AUROC (95% confidence interval 78-85%), 75% accuracy (70-79%), 52% sensitivity (41-62%), 92% specificity (86-98%), 85% PPV (76-95%), and 72% NPV (68-76%) on the internal dataset. External validation (18 cases) showed promising results: 94% AUROC (89-99%), 89% sensitivity (52-100%), 100% specificity (66-100%); in this small cohort, accuracy, sensitivity, and specificity did not differ significantly versus Node-RADS, with moderate agreement (Cohen κ = 0.47).

Conclusion: In this preliminary series, the model showed performance metrics in predicting LN involvement comparable to Node-RADS.

Relevance statement: Radiomics-based MRI models may represent a promising investigational tool for noninvasive axillary LN assessment in breast cancer. The performance comparable to Node-RADS suggests a potential to support clinical decision-making in the context of axillary de-escalation surgery.

Key points: Radiomics uses MRI to predict breast cancer LN involvement non-invasively and accurately. Radiomics and Node-RADS showed comparable performance. Radiomics could reduce invasive procedures, supporting personalized treatments in breast cancer care.

Background: Hormone receptor (HR) status guides breast cancer therapy. Deep learning (DL) applied to contrast-enhanced mammography (CEM) might offer a noninvasive means for HR status prediction, but class imbalance challenges model development and assessment. This preliminary study investigates CEM-based DL for HR status prediction, focusing on class imbalance handling.

Materials and methods: In this retrospective study, CEM tumor crops from 105 patients with invasive breast cancer were used. Patients were randomized into training (n = 68), validation (n = 16), and independent test (n = 21) sets. A "Residual Network 18" (ResNet-18), pretrained on ImageNet, was fine-tuned using weighted cross-entropy loss and an Adam optimizer. Model selection used validation area under the precision-recall curve (AUPRC); output probabilities were calibrated via temperature scaling. Performance was reported with accuracy, area under the receiver operating characteristic curve (AUROC), and imbalance-aware metrics (balanced accuracy, Matthews correlation coefficient (MCC)) with 95% confidence intervals (1,000-iteration bootstrap). Results are presented for standard (0.5) and optimized (validation F1-score for HR-negative class) thresholds.

Results: Validation AUPRC (model selection metric) was 0.640 (0.304-0.906). On the independent test set (optimized threshold 0.755), the model achieved 91.9% accuracy (86.5-97.3%), AUROC 0.808 (0.648-0.935), balanced accuracy 0.700 (0.550-0.853), and MCC 0.605 (0.296-0.818).

Conclusion: A ResNet-18, utilizing patient-level data splitting and imbalance-aware fine-tuning, can capture CEM features for HR status, performing well despite significant class imbalance. Generalizability is limited by dataset characteristics and acquisition specifics, warranting further validation in larger, diverse cohorts to establish clinical applicability.

Relevance statement: This work explores whether routinely acquired CEM images contain enough information for DL prediction of HR status. A ResNet-18 was trained with weighted loss and patient-level data splits; performance was quantified with imbalance-aware metrics to provide a realistic assessment in a highly skewed dataset, highlighting both the promise and current constraints of CEM-based molecular imaging.

Key points: A ResNet-18, optimized for class imbalance through weighted training and with calibrated probabilities, predicted HR positivity on CEM with 91.9% accuracy and AUROC 0.81 in an independent test cohort using an F1-tuned threshold. Balanced accuracy (0.70) and MCC (0.60) demonstrate maintained discrimination despite an approximate 85% class imbalance (HR-positive cases). Patient-level splitting was employed to ensure robust evaluation. Limitations related to the dataset's scope and specific imaging protocols may influence broader generalizability.

Background: We compared ultrahigh-resolution (UHR) photon-counting detector-computed tomography (PCD-CT) and spectral post-processed images for coronary stent visualization in a dynamic, anthropomorphic, and circulatory phantom.

Materials and methods: Ten coronary stents were scanned at 60, 80, and 100 beats per min (bpm) using UHR-spectral PCD-CT (96 × 0.2 mm collimation). Reconstructions included UHR (0.2 mm), downsampled (0.6 mm), and spectral post-processed images (0.4 mm), including virtual monoenergetic images (VMI; 45-100 keV), lumen-preserving images, and iodine maps (IM). Objective quality was assessed by measurable stent lumen visibility and stent strut width overestimation factor, compared to nominal strut width. Subjective quality was rated using a 4-point Likert scale. Repeated-measures analysis of variance-ANOVA and Friedman test with post hoc corrections were applied.

Results: UHR images provided the highest lumen visibility (62.6 ± 7.6%) at 60 bpm, outperforming all reconstructions ranging 43.5-52.7% (p ≤ 0.001) except IM (59.6 ± 11.9%, pairwise p = 0.839). UHR showed the lowest strut overestimation factor (18.1 ± 3.4), better than all spectral images (20.2-26.2, p ≤ 0.003) and DS (32.1 ± 6.5, p < 0.001). Subjective quality was best for UHR at 60 bpm (4.0 [interquartile range, IQR 4.0-4.0]) but declined at 100 bpm (3.0 [IQR 2.0-3.0], p < 0.01). VMI at 55 keV and IM maintained stable quality across heart rates (p ≥ 0.09).

Conclusion: PCD-CT combining UHR and spectral imaging enhances stent assessment. UHR provides the best lumen visibility and strut accuracy but suffers from motion artifacts, whereas VMIs at 55 keV and IM remain stable across heart rates and potentially provide incremental value.

Relevance statement: Combining UHR and spectral PCD-CT enhances coronary stent visualization by balancing high spatial detail with artifact reduction, potentially improving diagnostic confidence and enabling more reliable non-invasive follow-up across a range of heart rates in clinical practice.

Key points: The combined value of spectral UHR CT for stent imaging remains largely unexplored. UHR PCD-CT showed the highest lumen visibility and sharpest strut delineation, whilst being prone to motion artifacts. Spectral reconstructions complement UHR by reducing artifacts and stabilizing image quality, especially at higher heart rates.

von Hippel-Lindau (VHL) disease is a rare hereditary cancer syndrome caused by germline pathogenic variants in the VHL gene. The current standard of care primarily involves surgical resection, which is arbitrarily recommended for renal tumours ≥ 3 cm to reduce the risk of metastasis. However, this approach often leads to repeated surgeries and increased patient morbidity. The key unmet clinical need for VHL patients is the ability to predict the most appropriate therapeutic strategy and the optimal timing for surgical intervention on an individualised basis. Here, we describe a methodology designed to create an integrated map of intra- and inter-tumour heterogeneity in VHL-associated clear cell renal cell carcinoma by combining radiomics, histology, RNA sequencing, whole genome sequencing, and patient-derived organoid cultures from multi-regional tumour biopsies. We hypothesise that decoding this heterogeneity through an integrated analysis of imaging, histopathology, and molecular profiling will enhance diagnostic accuracy and enable more informed and personalised therapeutic decisions for VHL patients. RELEVANCE STATEMENT: Due to the current lack of biological or molecular markers assisting clinical decision-making, VHL patients undergo multiple surgical interventions with an incremental risk of complications and morbidity. We expect that our multimodal data study protocol will give tools to guide clinical management. KEY POINTS: Multiregional needle biopsies enable comprehensive analysis even in small ccRCC. Imaging characteristics suggest the presence of intra- and inter-lesion heterogeneity. Tumours are clonally independent and harbour distinct chromosome 3p loss events. Tumours display both intra- and inter-tumour transcriptomics heterogeneity. Patient-derived organoids grow more easily from areas of low tumour density.

Background: Patients with early-stage node-negative oral cancer undergo a sentinel lymph node biopsy (SLNB) or elective neck dissection under general anesthesia. A noninvasive imaging alternative would be of great interest. Superparamagnetic iron oxide (SPIO)-enhanced magnetic resonance imaging (MRI) can visualize draining lymph nodes and potentially metastases. We investigated the optimal combination of SPIO injection and T2*-weighted MRI settings to identify the sentinel lymph nodes (SLNs), the lymphatic drainage pattern, and possibly to detect metastatic SLNs.

Materials and methods: SPIO nanoparticles were injected submucosally around the primary tongue tumor in ten patients after routine SLNB imaging with indocyanine green-[^99mTc]Tc-nanocolloid, and MRI was performed 1 h after injection. SPIO dose was adjusted for every two patients based on the imaging quality. Drainage patterns were compared between single-photon emission computed tomography (SPECT)/computed tomography (CT) and MRI. MRI appearance of SLNs was compared to histopathology of resected nodes.

Results: One mg of iron was deemed a suitable dose after two dose alterations. All 25 SLNs observed on SPECT/CT in eight patients were also identified on MRI. Including higher echelons, 55 lymph nodes were seen on SPECT/CT, while 107 lymph nodes were seen on MRI. Eighteen lymph nodes showed a mixture of partial MRI signal attenuation and retention, which, when compared to histopathology, made discrimination between metastatic and nonmetastatic lymph nodes solely based on MRI impossible.

Conclusion: SPIO-enhanced T2*-weighted MRI is suitable for mapping SLNs and lymphatic drainage pattern, showing significantly more lymph nodes compared to SPECT/CT. Discriminating metastatic from nonmetastatic nodes does not seem feasible after SPIO injection.

Trial registration: Clinicaltrials.gov NCT04803331. Registered 4 March 2021; https://clinicaltrials.gov/study/NCT04803331 .

Relevance statement: SPIO-enhanced MRI seems noninferior to the current method of SLN detection with technetium, with better anatomical detail than SPECT/CT: if proven comparable, SPIO-enhanced MRI could be considered a nonradioactive alternative with higher spatial resolution to define lymphatic drainage of tumors.

Key points: The use of superparamagnetic iron oxide (SPIO)-enhanced MRI in head and neck cancer is underassessed. SPIO-enhanced MRI detects nodal drainage patterns comparably to SPECT/CT. SPIO-enhanced MRI does not discriminate lymph node metastases from false positives.

Background: Glioblastoma (GBM) is no longer regarded as a single disease, as distinct molecular subgroups exist, with the mesenchymal (MES) having the worst prognosis. As such, there is a critical need for noninvasive methods to determine GBM molecular status. Although conventional magnetic resonance imaging (MRI)-based radiomics showed promise for predicting GBM characteristics, few studies evaluated pipelines that leverage advanced diffusion MRI (dMRI) techniques, such as diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI), enabling characterization and quantification of tumor microstructure.

Materials and methods: To identify advanced dMRI radiomic features specific to MES GBM, we enrolled 36 GBM patients (4 mesenchymal, 32 non-mesenchymal), who underwent presurgical DTI and NODDI protocols. Post-surgery samples were processed to establish subgroup-specific GBM sphere-forming cell (GSC) lines, generating 21 xenografts (12 non-mesenchymal, 9 mesenchymal) that were subjected to the same dMRI protocols.

Results: By leveraging a preclinical-to-clinical transfer learning approach, a machine learning classification algorithm was developed to generalize between preclinical and clinical contexts. Models were trained on xenograft-derived data and validated using an independent patient test set. Using bootstrap resampling to estimate confidence intervals, the XGBoost model achieved an area under the receiver operating characteristic curve of 0.93 (95% confidence interval (CI): 0.79-1.00) and a balanced accuracy of 0.86 (0.64-1.00) for MES prediction. A subset of 9 selected features was sufficient to build a model that accurately predicted MES affiliation.

Conclusion: DTI and NODDI radiomics revealed key features that predict MES GBM and correlate with biological and clinical characteristics.

Relevance statement: A DTI and NODDI-based model trained on preclinical xenograft-derived data can be validated in a human patient cohort, demonstrating cross-species generalizability of radiomic biomarkers. This approach provides a noninvasive means to molecularly stratify GBM patients, enabling the potential to inform tailored treatment.

Key points: We defined a machine learning algorithm that, starting from subgroup-specific glioblastoma xenografts, reliably identifies the mesenchymal affiliation of glioblastoma patients. The specific dMRI features selected from experimental preclinical models of glioblastoma hold a remarkable predictive value. The same features provide insights into subgroup-restricted tumor tissue microstructure and its relationship with the malignant behavior of mesenchymal glioblastomas.

Background: Robust and continuous in vivo differentiation of spinal tracts along the brain-spinal cord axis is limited. We stitched brain and spinal cord diffusion tensor imaging (DTI) to create continuity between the brain and cervical spinal cord, enabling tractography along the central nervous system, producing an atlas of the spinal cord white matter.

Materials and methods: This prospective pilot study included four healthy subjects. Brain and cervical spinal cord 3-T DTI acquisitions were performed. Distortions were corrected using the Functional magnetic resonance imaging of the brain Software Library (FSL) software package. A semiautomatic stitching process was achieved using cross-correlation. Once the highest correlation peak was identified, rigid registration allowed accurate image alignment and fusion. Regions of interest were drawn in the brainstem according to atlas-guided projection tracts. Fiber tracking was performed using a deterministic approach with Diffusion Spectrum Imaging (DSI) Studio.

Results: The median fiber length from stitched-image tractography (192 mm) was significantly greater than that from both the brain (111.5 mm) and spinal cord (115 mm) fields of view. The white matter fiber atlas described: the corticospinal tract in the medial part of the lateral funiculus; the rubrospinal tract in the lateral funiculus, overlapped with the corticospinal tract; the gracilis and cuneatus tracts in the dorsal columns; the spinothalamic tract in the ventrolateral part of the spinal cord, around the ventral horn; and the spinocerebellar tracts overlapping them, in the lateral funiculus.

Conclusion: Stitching brain and spinal cord DTI fields of view provided an in vivo spinal cord white matter atlas in humans.

Relevance statement: This study provides a detailed and individualized mapping of spinal tracts, serving as a potential tool for neurosurgical planning, particularly in procedures involving intramedullary tumors. It also may enhance the accuracy of prognostic assessments in patients with spinal cord injury, multiple sclerosis, or degenerative myelopathy.

Key points: Brain and spinal cord diffusion tensor imaging scans were stitched to map spinal tracts across the central nervous system, enabling detailed three-dimensional imaging of spinal cord pathways. The resulting images showed precise locations of spinal tracts, producing an in vivo atlas of the spinal cord white matter. This tool may help surgeons plan safer operations and better predict outcomes in spinal cord disorders.