European Radiology Experimental最新文献

Background: We investigated the potential of an imaging-aware GPT-4-based chatbot in providing diagnoses based on imaging descriptions of abdominal pathologies.

Methods: Utilizing zero-shot learning via the LlamaIndex framework, GPT-4 was enhanced using the 96 documents from the Radiographics Top 10 Reading List on gastrointestinal imaging, creating a gastrointestinal imaging-aware chatbot (GIA-CB). To assess its diagnostic capability, 50 cases on a variety of abdominal pathologies were created, comprising radiological findings in fluoroscopy, MRI, and CT. We compared the GIA-CB to the generic GPT-4 chatbot (g-CB) in providing the primary and 2 additional differential diagnoses, using interpretations from senior-level radiologists as ground truth. The trustworthiness of the GIA-CB was evaluated by investigating the source documents as provided by the knowledge-retrieval mechanism. Mann-Whitney U test was employed.

Results: The GIA-CB demonstrated a high capability to identify the most appropriate differential diagnosis in 39/50 cases (78%), significantly surpassing the g-CB in 27/50 cases (54%) (p = 0.006). Notably, the GIA-CB offered the primary differential in the top 3 differential diagnoses in 45/50 cases (90%) versus g-CB with 37/50 cases (74%) (p = 0.022) and always with appropriate explanations. The median response time was 29.8 s for GIA-CB and 15.7 s for g-CB, and the mean cost per case was $0.15 and $0.02, respectively.

Conclusions: The GIA-CB not only provided an accurate diagnosis for gastrointestinal pathologies, but also direct access to source documents, providing insight into the decision-making process, a step towards trustworthy and explainable AI. Integrating context-specific data into AI models can support evidence-based clinical decision-making.

Relevance statement: A context-aware GPT-4 chatbot demonstrates high accuracy in providing differential diagnoses based on imaging descriptions, surpassing the generic GPT-4. It provided formulated rationale and source excerpts supporting the diagnoses, thus enhancing trustworthy decision-support.

Key points: • Knowledge retrieval enhances differential diagnoses in a gastrointestinal imaging-aware chatbot (GIA-CB). • GIA-CB outperformed the generic counterpart, providing formulated rationale and source excerpts. • GIA-CB has the potential to pave the way for AI-assisted decision support systems.

Background: This study investigates the potential of diffusion tensor imaging (DTI) in identifying penumbral volume (PV) compared to the standard gadolinium-required perfusion-diffusion mismatch (PDM), utilizing a stack-based ensemble machine learning (ML) approach with enhanced explainability.

Methods: Sixteen male rats were subjected to middle cerebral artery occlusion. The penumbra was identified using PDM at 30 and 90 min after occlusion. We used 11 DTI-derived metrics and 14 distance-based features to train five voxel-wise ML models. The model predictions were integrated using stack-based ensemble techniques. ML-estimated and PDM-defined PVs were compared to evaluate model performance through volume similarity assessment, the Pearson correlation analysis, and Bland-Altman analysis. Feature importance was determined for explainability.

Results: In the test rats, the ML-estimated median PV was 106.4 mL (interquartile range 44.6-157.3 mL), whereas the PDM-defined median PV was 102.0 mL (52.1-144.9 mL). These PVs had a volume similarity of 0.88 (0.79-0.96), a Pearson correlation coefficient of 0.93 (p < 0.001), and a Bland-Altman bias of 2.5 mL (2.4% of the mean PDM-defined PV), with 95% limits of agreement ranging from -44.9 to 49.9 mL. Among the features used for PV prediction, the mean diffusivity was the most important feature.

Conclusions: Our study confirmed that PV can be estimated using DTI metrics with a stack-based ensemble ML approach, yielding results comparable to the volume defined by the standard PDM. The model explainability enhanced its clinical relevance. Human studies are warranted to validate our findings.

Relevance statement: The proposed DTI-based ML model can estimate PV without the need for contrast agent administration, offering a valuable option for patients with kidney dysfunction. It also can serve as an alternative if perfusion map interpretation fails in the clinical setting.

Key points: • Penumbral volume can be estimated by DTI combined with stack-based ensemble ML. • Mean diffusivity was the most important feature used for predicting penumbral volume. • The proposed approach can be beneficial for patients with kidney dysfunction.

Overall quality of radiomics research has been reported as low in literature, which constitutes a major challenge to improve. Consistent, transparent, and accurate reporting is critical, which can be accomplished with systematic use of reporting guidelines. The CheckList for EvaluAtion of Radiomics research (CLEAR) was previously developed to assist authors in reporting their radiomic research and to assist reviewers in their evaluation. To take full advantage of CLEAR, further explanation and elaboration of each item, as well as literature examples, may be useful. The main goal of this work, Explanation and Elaboration with Examples for CLEAR (CLEAR-E3), is to improve CLEAR's usability and dissemination. In this international collaborative effort, members of the European Society of Medical Imaging Informatics-Radiomics Auditing Group searched radiomics literature to identify representative reporting examples for each CLEAR item. At least two examples, demonstrating optimal reporting, were presented for each item. All examples were selected from open-access articles, allowing users to easily consult the corresponding full-text articles. In addition to these, each CLEAR item's explanation was further expanded and elaborated. For easier access, the resulting document is available at https://radiomic.github.io/CLEAR-E3/ . As a complementary effort to CLEAR, we anticipate that this initiative will assist authors in reporting their radiomics research with greater ease and transparency, as well as editors and reviewers in reviewing manuscripts.Relevance statement Along with the original CLEAR checklist, CLEAR-E3 is expected to provide a more in-depth understanding of the CLEAR items, as well as concrete examples for reporting and evaluating radiomic research.Key points• As a complementary effort to CLEAR, this international collaborative effort aims to assist authors in reporting their radiomics research, as well as editors and reviewers in reviewing radiomics manuscripts.• Based on positive examples from the literature selected by the EuSoMII Radiomics Auditing Group, each CLEAR item explanation was further elaborated in CLEAR-E3.• The resulting explanation and elaboration document with examples can be accessed at  https://radiomic.github.io/CLEAR-E3/ .

Background: Chondrosarcomas are rare malignant bone tumors diagnosed by analyzing radiological images and histology of tissue biopsies and evaluating features such as matrix calcification, cortical destruction, trabecular penetration, and tumor cell entrapment.

Methods: We retrospectively analyzed 16 cartilaginous tumor tissue samples from three patients (51-, 54-, and 70-year-old) diagnosed with a dedifferentiated chondrosarcoma at the femur, a moderately differentiated chondrosarcoma in the pelvis, and a predominantly moderately differentiated chondrosarcoma at the scapula, respectively. We combined a hematein-based x-ray staining with high-resolution three-dimensional (3D) microscopic x-ray computed tomography (micro-CT) for nondestructive 3D tumor assessment and tumor margin evaluation.

Results: We detected trabecular entrapment on 3D micro-CT images and followed bone destruction throughout the volume. In addition to staining cell nuclei, hematein-based staining also improved the visualization of the tumor matrix, allowing for the distinction between the tumor and the bone marrow cavity. The hematein-based staining did not interfere with further conventional histology. There was a 5.97 ± 7.17% difference between the relative tumor area measured using micro-CT and histopathology (p = 0.806) (Pearson correlation coefficient r = 0.92, p = 0.009). Signal intensity in the tumor matrix (4.85 ± 2.94) was significantly higher in the stained samples compared to the unstained counterparts (1.92 ± 0.11, p = 0.002).

Conclusions: Using nondestructive 3D micro-CT, the simultaneous visualization of radiological and histopathological features is feasible.

Relevance statement: 3D micro-CT data supports modern radiological and histopathological investigations of human bone tumor specimens. It has the potential for being an integrative part of clinical preoperative diagnostics.

Key points: • Matrix calcifications are a relevant diagnostic feature of bone tumors. • Micro-CT detects all clinically diagnostic relevant features of x-ray-stained chondrosarcoma. • Micro-CT has the potential to be an integrative part of clinical diagnostics.

Background: We compared computed tomography (CT) images and holograms (HG) to assess the number of arteries of the lung lobes undergoing lobectomy and assessed easiness in interpretation by radiologists and thoracic surgeons with both techniques.

Methods: Patients scheduled for lobectomy for lung cancer were prospectively included and underwent CT for staging. A patient-specific three-dimensional model was generated and visualized in an augmented reality setting. One radiologist and one thoracic surgeon evaluated CT images and holograms to count lobar arteries, having as reference standard the number of arteries recorded at surgery. The easiness of vessel identification was graded according to a Likert scale. Wilcoxon signed-rank test and κ statistics were used.

Results: Fifty-two patients were prospectively included. The two doctors detected the same number of arteries in 44/52 images (85%) and in 51/52 holograms (98%). The mean difference between the number of artery branches detected by surgery and CT images was 0.31 ± 0.98, whereas it was 0.09 ± 0.37 between surgery and HGs (p = 0.433). In particular, the mean difference in the number of arteries detected in the upper lobes was 0.67 ± 1.08 between surgery and CT images and 0.17 ± 0.46 between surgery and holograms (p = 0.029). Both radiologist and surgeon showed a higher agreement for holograms (κ = 0.99) than for CT (κ = 0.81) and found holograms easier to evaluate than CTs (p < 0.001).

Conclusions: Augmented reality by holograms is an effective tool for preoperative vascular anatomy assessment of lungs, especially when evaluating the upper lobes, more prone to anatomical variations.

Trial registration: ClinicalTrials.gov, NCT04227444 RELEVANCE STATEMENT: Preoperative evaluation of the lung lobe arteries through augmented reality may help the thoracic surgeons to carefully plan a lobectomy, thus contributing to optimize patients' outcomes.

Key points: • Preoperative assessment of the lung arteries may help surgical planning. • Lung artery detection by augmented reality was more accurate than that by CT images, particularly for the upper lobes. • The assessment of the lung arterial vessels was easier by using holograms than CT images.

Objective: Guyon's canal syndrome is caused by compression of the ulnar nerve at the wrist, occasionally requiring decompression surgery. In recent times, minimally invasive approaches have gained popularity. The aim of this study was to assess the efficacy and safety of ultrasound-guided thread release for transecting the palmar ligament in Guyon's canal without harming surrounding structures, in a cadaveric specimen model.

Methods: After ethical approval, thirteen ultrasound-guided thread releases of Guyon's canal were performed on the wrists of softly embalmed anatomic specimens. Cadavers showing injuries or prior operations at the hand were excluded. Subsequently, the specimens were dissected, and the outcome of the interventions and potential damage to adjacent anatomical structures as well as ultrasound visibility were evaluated with a score from one to three.

Results: Out of 13 interventions, a complete transection was achieved in ten cases (76.9%), and a partial transection was documented in three cases (23.1%). Irrelevant lesions on the flexor tendons were observed in two cases (15.4%), and an arterial branch was damaged in one (7.7%). Ultrasound visibility varied among specimens, but essential structures were delineated in all cases.

Conclusion: Ultrasound-guided thread release of Guyon's canal has shown promising first results in anatomic specimens. However, further studies are required to ensure the safety of the procedure.

Relevance statement: Our study showed that minimally invasive ultrasound-guided thread release of Guyon's canal is a feasible approach in the anatomical model. The results may provide a basis for further research and refinement of this technique.

Key points: • In Guyon's canal syndrome, the ulnar nerve is compressed at the wrist, often requiring surgical release. • We adapted and tested a minimally invasive ultrasound-guided thread release technique in anatomic specimens. • The technique was effective; however, in one specimen, a small anatomic branch was damaged.

Background: To evaluate the reproducibility of a vessel-specific minimum cost path (MCP) technique used for lobar segmentation on noncontrast computed tomography (CT).

Methods: Sixteen Yorkshire swine (49.9 ± 4.7 kg, mean ± standard deviation) underwent a total of 46 noncontrast helical CT scans from November 2020 to May 2022 using a 320-slice scanner. A semiautomatic algorithm was employed by three readers to segment the lung tissue and pulmonary arterial tree. The centerline of the arterial tree was extracted and partitioned into six subtrees for lobar assignment. The MCP technique was implemented to assign lobar territories by assigning lung tissue voxels to the nearest arterial tree segment. MCP-derived lobar mass and volume were then compared between two acquisitions, using linear regression, root mean square error (RMSE), and paired sample t-tests. An interobserver and intraobserver analysis of the lobar measurements was also performed.

Results: The average whole lung mass and volume was 663.7 ± 103.7 g and 1,444.22 ± 309.1 mL, respectively. The lobar mass measurements from the initial (MLobe1) and subsequent (MLobe2) acquisitions were correlated by MLobe1 = 0.99 MLobe2 + 1.76 (r = 0.99, p = 0.120, RMSE = 7.99 g). The lobar volume measurements from the initial (VLobe1) and subsequent (VLobe2) acquisitions were correlated by VLobe1 = 0.98VLobe2 + 2.66 (r = 0.99, p = 0.160, RSME = 15.26 mL).

Conclusions: The lobar mass and volume measurements showed excellent reproducibility through a vessel-specific assignment technique. This technique may serve for automated lung lobar segmentation, facilitating clinical regional pulmonary analysis.

Relevance statement: Assessment of lobar mass or volume in the lung lobes using noncontrast CT may allow for efficient region-specific treatment strategies for diseases such as pulmonary embolism and chronic thromboembolic pulmonary hypertension.

Key points: • Lobar segmentation is essential for precise disease assessment and treatment planning. • Current methods for segmentation using fissure lines are problematic. • The minimum-cost-path technique here is proposed and a swine model showed excellent reproducibility for lobar mass measurements. • Interobserver agreement was excellent, with intraclass correlation coefficients greater than 0.90.

Background: We aimed to improve the image quality (IQ) of sparse-view computed tomography (CT) images using a U-Net for lung metastasis detection and determine the best tradeoff between number of views, IQ, and diagnostic confidence.

Methods: CT images from 41 subjects aged 62.8 ± 10.6 years (mean ± standard deviation, 23 men), 34 with lung metastasis, 7 healthy, were retrospectively selected (2016-2018) and forward projected onto 2,048-view sinograms. Six corresponding sparse-view CT data subsets at varying levels of undersampling were reconstructed from sinograms using filtered backprojection with 16, 32, 64, 128, 256, and 512 views. A dual-frame U-Net was trained and evaluated for each subsampling level on 8,658 images from 22 diseased subjects. A representative image per scan was selected from 19 subjects (12 diseased, 7 healthy) for a single-blinded multireader study. These slices, for all levels of subsampling, with and without U-Net postprocessing, were presented to three readers. IQ and diagnostic confidence were ranked using predefined scales. Subjective nodule segmentation was evaluated using sensitivity and Dice similarity coefficient (DSC); clustered Wilcoxon signed-rank test was used.

Results: The 64-projection sparse-view images resulted in 0.89 sensitivity and 0.81 DSC, while their counterparts, postprocessed with the U-Net, had improved metrics (0.94 sensitivity and 0.85 DSC) (p = 0.400). Fewer views led to insufficient IQ for diagnosis. For increased views, no substantial discrepancies were noted between sparse-view and postprocessed images.

Conclusions: Projection views can be reduced from 2,048 to 64 while maintaining IQ and the confidence of the radiologists on a satisfactory level.

Relevance statement: Our reader study demonstrates the benefit of U-Net postprocessing for regular CT screenings of patients with lung metastasis to increase the IQ and diagnostic confidence while reducing the dose.

Key points: • Sparse-projection-view streak artifacts reduce the quality and usability of sparse-view CT images. • U-Net-based postprocessing removes sparse-view artifacts while maintaining diagnostically accurate IQ. • Postprocessed sparse-view CTs drastically increase radiologists' confidence in diagnosing lung metastasis.

Artificial intelligence (AI) has demonstrated great potential in a wide variety of applications in interventional radiology (IR). Support for decision-making and outcome prediction, new functions and improvements in fluoroscopy, ultrasound, computed tomography, and magnetic resonance imaging, specifically in the field of IR, have all been investigated. Furthermore, AI represents a significant boost for fusion imaging and simulated reality, robotics, touchless software interactions, and virtual biopsy. The procedural nature, heterogeneity, and lack of standardisation slow down the process of adoption of AI in IR. Research in AI is in its early stages as current literature is based on pilot or proof of concept studies. The full range of possibilities is yet to be explored.

Relevance statement Exploring AI’s transformative potential, this article assesses its current applications and challenges in IR, offering insights into decision support and outcome prediction, imaging enhancements, robotics, and touchless interactions, shaping the future of patient care.

Key points

• AI adoption in IR is more complex compared to diagnostic radiology.

• Current literature about AI in IR is in its early stages.

• AI has the potential to revolutionise every aspect of IR.

Graphical Abstract

Background

To compare denoising diffusion probabilistic models (DDPM) and generative adversarial networks (GAN) for recovering contrast-enhanced breast magnetic resonance imaging (MRI) subtraction images from virtual low-dose subtraction images.

Methods

Retrospective, ethically approved study. DDPM- and GAN-reconstructed single-slice subtraction images of 50 breasts with enhancing lesions were compared to original ones at three dose levels (25%, 10%, 5%) using quantitative measures and radiologic evaluations. Two radiologists stated their preference based on the reconstruction quality and scored the lesion conspicuity as compared to the original, blinded to the model. Fifty lesion-free maximum intensity projections were evaluated for the presence of false-positives. Results were compared between models and dose levels, using generalized linear mixed models.

Results

At 5% dose, both radiologists preferred the GAN-generated images, whereas at 25% dose, both radiologists preferred the DDPM-generated images. Median lesion conspicuity scores did not differ between GAN and DDPM at 25% dose (5 versus 5, p = 1.000) and 10% dose (4 versus 4, p = 1.000). At 5% dose, both readers assigned higher conspicuity to the GAN than to the DDPM (3 versus 2, p = 0.007). In the lesion-free examinations, DDPM and GAN showed no differences in the false-positive rate at 5% (15% versus 22%), 10% (10% versus 6%), and 25% (6% versus 4%) (p = 1.000).

Conclusions

Both GAN and DDPM yielded promising results in low-dose image reconstruction. However, neither of them showed superior results over the other model for all dose levels and evaluation metrics. Further development is needed to counteract false-positives.

Relevance statement

For MRI-based breast cancer screening, reducing the contrast agent dose is desirable. Diffusion probabilistic models and generative adversarial networks were capable of retrospectively enhancing the signal of low-dose images. Hence, they may supplement imaging with reduced doses in the future.

Key points

• Deep learning may help recover signal in low-dose contrast-enhanced breast MRI.

• Two models (DDPM and GAN) were trained at different dose levels.

• Radiologists preferred DDPM at 25%, and GAN images at 5% dose.

• Lesion conspicuity between DDPM and GAN was similar, except at 5% dose.

• GAN and DDPM yield promising results in low-dose image reconstruction.

Graphical Abstract