European Radiology Experimental最新文献

Background: Transarterial radioembolisation (TARE) is a treatment for liver malignancies, involving the injection of radioactive microspheres in the hepatic artery (HA). Tumour-to-nontumour uptake varies among patients, possibly influenced by patient-specific blood flow profiles. To examine the impact of HA blood flow rate and high microsphere dosages on microsphere distribution in normal liver parenchyma, ex vivo magnetic resonance imaging (MRI)-guided machine perfusion experiments were conducted in porcine livers.

Materials and methods: Porcine livers were subjected to oxygenated normothermic machine perfusion at three HA flow rates (0.02, 0.15, and 0.22 mL/min/g liver tissue; n = 3 per condition). Five fractions of 250 mg nonradioactive ¹⁶⁵Ho-loaded microspheres were administered to n = 9 livers, and four additional fractions of 1,000 mg to n = 6 livers. Dynamic contrast-enhanced and Ho-sensitive T2*-weighed MR scans were acquired to extract perfusion rates, fictive dose maps, and homogeneity indices (HI).

Results: Microsphere distribution correlated moderately with perfusion rate at low HA flow rate (r = 0.611), and very strongly at higher HA flow rates (r = 0.977 and 0.951 for 0.15 and 0.22 mL/min/g, respectively). Homogeneity increased with increasing flow rates, with HIs ranging from 3.68-4.72 at low, to 2.01-2.66 at medium, and 1.60-2.36 at high HA flow rate. HI decreased with higher microsphere concentrations, though distribution patterns remained unchanged.

Conclusion: In our ex vivo model, higher HA flow rates resulted in more homogeneous microsphere distributions. The impact on tumourous tissue needs further investigation to determine whether pre-TARE HA blood flow measurements could improve microsphere distribution predictions.

Relevance statement: Mapping of the hepatic arterial blood flow rate before transarterial radioembolisation and adjusting the treatment accordingly may help to improve outcomes for patients with liver cancer.

Key points: Parameters influencing microsphere distribution were studied in MRI-perfused healthy porcine livers. Higher hepatic arterial blood flow rates led to more homogeneous microsphere distributions. Administering large numbers of microspheres did not alter microsphere distribution patterns. Impact on tumour tissue should be further investigated.

Background: Detecting malignant transformation of sinonasal inverted papilloma (SIP) into squamous cell carcinoma (SIP-SCC) before surgery is a clinical need. We aimed to explore the value of deep learning (DL) that leverages nasal endoscopy and T2-weighted magnetic resonance imaging (T2W-MRI) for automated tumor segmentation and differentiation between SIP and SIP-SCC.

Methods: We conducted a retrospective analysis of 174 patients diagnosed with SIPs, who were divided into a training cohort (n = 121) and a testing cohort (n = 53). Three DL architectures were utilized to train automated segmentation models for endoscopic and T2W-MRI images. DL scores predicting SIP-SCC were generated using DenseNet121 from both modalities and combined to create a dual-modality DL nomogram. The diagnostic performance of the DL models was assessed alongside two radiologists, evaluated through the area under the receiver operating characteristic curve (AUROC), with comparisons made using the Delong method.

Results: In the testing cohort, the FCN_ResNet101 and VNet exhibited superior performance in automated segmentation, achieving mean dice similarity coefficients of 0.95 ± 0.03 for endoscopy and 0.93 ± 0.02 for T2W-MRI, respectively. The dual-modality DL nomogram based on automated segmentation demonstrated the highest predictive performance for SIP-SCC (AUROC 0.865), outperforming the radiology resident (AUROC 0.672, p = 0.071) and the attending radiologist (AUROC 0.707, p = 0.066), with a trend toward significance. Notably, both radiologists improved their diagnostic performance with the assistance of the DL nomogram (AUROCs 0.734 and 0.834).

Conclusion: The DL framework integrating endoscopy and T2W-MRI offers a fully automated predictive tool for SIP-SCC.

Relevance statement: The integration of endoscopy and T2W-MRI within a well-established DL framework enables fully automated prediction of SIP-SSC, potentially improving decision-making for patients with suspicious SIP.

Key points: Detecting the transformation of SIP into SIP-SCC before surgery is both critical and challenging. Endoscopy and T2W-MRI were integrated using DL for predicting SIP-SCC. The dual-modality DL nomogram outperformed two radiologists. The nomogram may improve decision-making for patients with suspicious SIP.

Severity scores, which often refer to the likelihood or probability of a pathology, are commonly provided by artificial intelligence (AI) tools in radiology. However, little attention has been given to the use of these AI scores, and there is a lack of transparency into how they are generated. In this comment, we draw on key principles from psychological science and statistics to elucidate six human factors limitations of AI scores that undermine their utility: (1) variability across AI systems; (2) variability within AI systems; (3) variability between radiologists; (4) variability within radiologists; (5) unknown distribution of AI scores; and (6) perceptual challenges. We hypothesize that these limitations can be mitigated by providing the false discovery rate and false omission rate for each score as a threshold. We discuss how this hypothesis could be empirically tested. KEY POINTS: The radiologist-AI interaction has not been given sufficient attention. The utility of AI scores is limited by six key human factors limitations. We propose a hypothesis for how to mitigate these limitations by using false discovery rate and false omission rate.

Background: Differentiation between saltwater and freshwater immersion as well as estimating the corpse's time in water can be challenging. We aimed to establish and examine the feasibility of a novel approach based on sodium magnetic resonance imaging (²³Na MRI) of the eye to facilitate noninvasive sodium quantification.

Methods: Enucleated porcine eyes were immersed in NaCl 0.9%, NaCl 3.0%, NaCl 5.85%, distilled water (DW) or lake water (LW) at different time intervals, followed by ²³Na 7-T MRI sodium quantification.

Results: After 6 h of immersion, a significant difference in vitreous body (VB) sodium concentration was found for NaCl 5.85% versus DW or LW (p ≤ 0.019). After 24 and 48 h of immersion, a significant difference in VB sodium concentration was found for NaCl 5.85% versus DW, LW, NaCl 3.0% or NaCl 0.9%, as well as for NaCl 3.0% versus DW, LW or NaCl 0.9% (p ≤ 0.001). After 24 h of immersion, lens sodium concentration showed a significant difference for NaCl 5.85% versus DW, LW, NaCl 3.0% or NaCl 0.9% (p ≤ 0.009); after 48 h of immersion, for NaCl 5.85% versus DW, LW, NaCl 3.0% or NaCl 0.9% (p ≤ 0.001), as well as for NaCl 3.0% versus DW, LW or NaCl 0.9% (p ≤ 0.007). For VB, sodium concentration changes over immersion time, and exponential curves were fitted to the data.

Conclusion: Using ²³Na MRI in ex vivo porcine eyes with different immersion times in various saltwater concentrations and freshwater equivalents allowed noninvasive quantification of VB and lens sodium concentrations.

Relevance statement: Although not a substitute for autopsy, ²³Na MRI assessment of VB and lens sodium concentrations may provide biochemical support in suspected drowning, especially in cases where an internal examination of the body is not authorized or where objections to autopsy are upheld.

Key points: Postmortem porcine eyes with different immersion times in saltwater and freshwater. Noninvasive quantification of vitreous body and lens sodium concentrations with ²³Na MRI. Exponential time course of vitreous body sodium concentration in saltwater and freshwater.

Objective: To evaluate in vivo a fully integrated photoacoustic tomography imaging system based on Fabry-Pérot ultrasound sensing method applied on porcine abdominal organs. This approach could be used by surgeons during intraoperative clinical procedures.

Methods: The photoacoustic imaging system was fully integrated into a single structure, and the detection technology was based on a Fabry-Pérot interferometer. The detection probe connected to the imaging system was applied directly to the organs of a male "large white" Sus scrofa pig weighing 80 kg, either manually or using a stand, with or without a gel interface. All experiments were performed in compliance with EU Directive 2010/63/EU on animal experimentation (APAFiS #31507).

Results: All intraperitoneal and retroperitoneal organs were evaluated using photoacoustic imaging. The evaluation of both hollow and solid organs was successfully conducted with consistent three-dimensional image quality. We demonstrate the system's ability to image blood vessels with diameters ranging from several millimeters down to less than 100 µm. Macroscopic evaluation of the organs using photoacoustic tomography imaging did not reveal any damage or burns caused by the excitation laser.

Conclusion: To our knowledge, this is the first reported imaging session of abdominal organs in an in vivo porcine model, performed using a photoacoustic tomography system with Fabry-Pérot interferometer detection. We present a high-resolution photoacoustic tomography system that is closer to routine clinical translation, thanks to a fully integrated system.

Relevance statement: Photoacoustic evaluation of organs using a fully integrated system could become a valuable tool for surgical teams for intraprocedural assessment of vascularization.

Key points: Photoacoustic imaging visualizes blood vessels without contrast agents or ionizing radiation. Photoacoustic imaging systems detect blood vessels ranging from millimeters to 100 µm. Fully integrated photoacoustic imaging systems are autonomously operable by surgical teams.

Background: Aging alters musculoskeletal structure and function, affecting muscle mass, composition, and strength, increasing the risk of falls and loss of independence in older adults. This study assessed cross-sectional area (CSA) and fat infiltration (FI) of six thigh muscles through a validated deep learning model. Gender differences and correlations between fat, muscle parameters, and age were also analyzed.

Methods: We retrospectively analyzed 141 participants (67 females, 74 males) aged 52-82 years. Participants underwent magnetic resonance imaging (MRI) scans of the right thigh and dual-energy x-ray absorptiometry to determine appendicular skeletal muscle mass index (ASMMI) and body fat percentage (FAT%). A deep learning-based application was developed to automate the segmentation of six thigh muscle groups.

Results: Deep learning model accuracy was evaluated using the "intersection over union" (IoU) metric, with average IoU values across muscle groups ranging from 0.84 to 0.99. Mean CSA was 10,766.9 mm² (females 8,892.6 mm², males 12,463.9 mm², p < 0.001). The mean FI value was 14.92% (females 17.42%, males 12.62%, p < 0.001). Males showed larger CSA and lower FI in all thigh muscles compared to females. Positive correlations were identified in females between the FI of posterior thigh muscle groups (biceps femoris, semimembranosus, and semitendinosus) and age (r or ρ = 0.35-0.48; p ≤ 0.004), while no significant correlations were observed between CSA, ASMMI, or FAT% and age.

Conclusion: Deep learning accurately quantifies muscle CSA and FI, reducing analysis time and human error. Aging impacts on muscle composition and distribution and gender-specific assessments in older adults is needed.

Relevance statement: Efficient deep learning-based MRI image segmentation to assess the composition of six thigh muscle groups in over 50 individuals revealed gender differences in thigh muscle CSA and FI. These findings have potential clinical applications in assessing muscle quality, decline, and frailty.

Key points: Deep learning model enhanced MRI segmentation, providing high assessment accuracy. Significant gender differences in cross-sectional area and fat infiltration across all thigh muscles were observed. In females, fat infiltration of the posterior thigh muscles was positively correlated with age.

Background: To determine changes in quantitative T1 relaxation times (qT1) in deep gray matter in patients recovered from coronavirus disease 2019 (COVID-19).

Methods: Unvaccinated COVID-19 participants ≥ 3 months after seropositivity and age- and sex-matched controls were examined using 3-T magnetic resonance imaging. Bilateral measures of thalamus, pallidum, putamen, caudate and accumbens nuclei, and hippocampus were extracted from qT1 maps after automated segmentation. Baseline characteristics and results of tests assessing neurological functions (standardized exam), ability to smell (4-Item Pocket Smell Test), depression (Beck Depression Inventory-II), sleepiness (Epworth Sleepiness Scale), sleep quality (Pittsburgh Sleep Quality Index), health-related quality of life (EQ-5D), and cognitive performance (Montreal Cognitive Assessment) were evaluated.

Results: One hundred forty-five subjects (median age, 46 years; 73 females) were included (11/2020-12/2021): 69 recovered after COVID-19 and 76 controls (age, p = 0.532; sex, p = 0.799), without significant differences in qT1 values overall (all p-values > 0.050). Subgroup analysis of participants aged ≥ 40 (age, p = 0.675; sex, p = 0.447) revealed higher qT1 values in previously hospitalized COVID-19 subjects (23/69) compared to controls (47/76) in left and right caudate nuclei (p = 0.009; p = 0.027), left accumbens nucleus (p = 0.017), right putamen (p = 0.041), and right hippocampus (p = 0.020). No correlations were found with macroscopic imaging findings, pre-existing conditions, time since COVID-19 diagnosis, inpatient treatment duration, or test results.

Conclusion: T1 mapping revealed microstructural changes in striatal and hippocampal regions of unvaccinated individuals aged ≥ 40 who recovered from moderate-to-severe COVID-19 during the pre-Omicron era.

Relevance statement: This study elucidates brain involvement following severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, underscoring the need for further longitudinal analyses to assess the potential reversibility, stability or deterioration of these findings.

Key points: We hypothesized altered T1 relaxation times in deep gray matter after COVID-19. Unvaccinated participants ≥ 40 years exhibited higher striatal, hippocampal qT1 after moderate-to-severe COVID-19. No qT1 correlations were found with hospitalization duration, pre-existing conditions, or neuro-(psycho)logical tests.

Osteoporosis is widespread with a high incidence rate, resulting in fragility fractures which are a major contributor to mortality among the elderly. Artificial intelligence (AI), in particular artificial neural networks, appears to be useful in managing osteoporosis complexity, where bone mineral density usually reduces with aging, losing the pivotal role in decision-making regarding fracture prediction and treatment choice. Nevertheless, only some osteoporotic patients develop fragility fractures, and treatments often are not prescribed because of the high costs and poor patient adherence. AI can help clinicians to identify patients prone to fragility fractures who can benefit from preventive interventions. We describe herein the methodology issues underlying the potential advantages of introducing AI methods to support clinical decision-making in osteoporosis, being aware of challenges regarding data availability and quality, model interpretability, integration into clinical workflows, and validation of predictive accuracy. The fact that no AI fracture risk prediction software is still publicly available can be related to the fact that few high-quality datasets are available and that AI models, particularly deep learning approaches, often act as 'black boxes', making it difficult to understand how predictions are made. In addition, the effective implementation of predictive software has not reached sufficient integration with existing systems. RELEVANCE STATEMENT: With aging, bone mineral density may lose the pivotal role in osteoporosis decision-making regarding fracture prediction and treatment choice. In this scenario, AI, particularly artificial neural networks (ANNs), can be useful in supporting the clinical management of patients affected by osteoporosis. KEY POINTS: Osteoporosis is a complex disease with many interlinked clinical and radiological variables. Bone mineral density and other known indices do not allow optimal decision-making in patients affected by osteoporosis. ANN analysis can better discriminate osteoporotic patients particularly prone to fragility fractures and can predict future fractures.

Background: In acute neck infections, magnetic resonance imaging (MRI) shows retropharyngeal edema (RPE), which is a prognostic imaging biomarker for a severe course of illness. This study aimed to develop a deep learning-based algorithm for the automated detection of RPE.

Methods: We developed a deep neural network consisting of two parts using axial T2-weighted water-only Dixon MRI images from 479 patients with acute neck infections annotated by radiologists at both slice and patient levels. First, a convolutional neural network (CNN) classified individual slices; second, an algorithm classified patients based on a stack of slices. Model performance was compared with the radiologists' assessment as a reference standard. Accuracy, sensitivity, specificity, and area under receiver operating characteristic curve (AUROC) were calculated. The proposed CNN was compared with InceptionV3, and the patient-level classification algorithm was compared with traditional machine learning models.

Results: Of the 479 patients, 244 (51%) were positive and 235 (49%) negative for RPE. Our model achieved accuracy, sensitivity, specificity, and AUROC of 94.6%, 83.3%, 96.2%, and 94.1% at the slice level, and 87.4%, 86.5%, 88.2%, and 94.8% at the patient level, respectively. The proposed CNN was faster than InceptionV3 but equally accurate. Our patient classification algorithm outperformed traditional machine learning models.

Conclusion: A deep learning model, based on weakly annotated data and computationally manageable training, achieved high accuracy for automatically detecting RPE on MRI in patients with acute neck infections.

Relevance statement: Our automated method for detecting relevant MRI findings was efficiently trained and might be easily deployed in practice to study clinical applicability. This approach might improve early detection of patients at high risk for a severe course of acute neck infections.

Key points: Deep learning automatically detected retropharyngeal edema on MRI in acute neck infections. Areas under the receiver operating characteristic curve were 94.1% at the slice level and 94.8% at the patient level. The proposed convolutional neural network was lightweight and required only weakly annotated data.

Background: To evaluate the impact of an annotation guideline on the performance of large language models (LLMs) in extracting data from stroke computed tomography (CT) reports.

Methods: The performance of GPT-4o and Llama-3.3-70B in extracting ten imaging findings from stroke CT reports was assessed in two datasets from a single academic stroke center. Dataset A (n = 200) was a stratified cohort including various pathological findings, whereas dataset B (n = 100) was a consecutive cohort. Initially, an annotation guideline providing clear data extraction instructions was designed based on a review of cases with inter-annotator disagreements in dataset A. For each LLM, data extraction was performed under two conditions: with the annotation guideline included in the prompt and without it.

Results: GPT-4o consistently demonstrated superior performance over Llama-3.3-70B under identical conditions, with micro-averaged precision ranging from 0.83 to 0.95 for GPT-4o and from 0.65 to 0.86 for Llama-3.3-70B. Across both models and both datasets, incorporating the annotation guideline into the LLM input resulted in higher precision rates, while recall rates largely remained stable. In dataset B, the precision of GPT-4o and Llama-3-70B improved from 0.83 to 0.95 and from 0.87 to 0.94, respectively. Overall classification performance with and without the annotation guideline was significantly different in five out of six conditions.

Conclusion: GPT-4o and Llama-3.3-70B show promising performance in extracting imaging findings from stroke CT reports, although GPT-4o steadily outperformed Llama-3.3-70B. We also provide evidence that well-defined annotation guidelines can enhance LLM data extraction accuracy.

Relevance statement: Annotation guidelines can improve the accuracy of LLMs in extracting findings from radiological reports, potentially optimizing data extraction for specific downstream applications.

Key points: LLMs have utility in data extraction from radiology reports, but the role of annotation guidelines remains underexplored. Data extraction accuracy from stroke CT reports by GPT-4o and Llama-3.3-70B improved when well-defined annotation guidelines were incorporated into the model prompt. Well-defined annotation guidelines can improve the accuracy of LLMs in extracting imaging findings from radiological reports.