Background: Artificial intelligence (AI) and deep learning are increasingly applied in cardiovascular imaging. However, the “black box” nature of these models raises challenges for clinical trust and integration. Explainable Artificial Intelligence (XAI) seeks to address these concerns by providing insights into model decision-making. This systematic review synthesizes current research on the use of XAI methods in radiological cardiovascular imaging. Methods: A systematic literature search was conducted in PubMed, Scopus, and Web of Science to identify peer-reviewed original research articles published between January 2015 and March 2025. Studies were included if they applied XAI techniques—such as Gradient-Weighted Class Activation Mapping (Grad-CAM), Shapley Additive Explanations (SHAPs), Local Interpretable Model-Agnostic Explanations (LIMEs), or saliency maps—to cardiovascular imaging modalities, including cardiac computed tomography (CT), magnetic resonance imaging (MRI), echocardiography and other ultrasound examinations, and chest X-ray (CXR). Studies focusing on nuclear medicine, structured/tabular data without imaging, or lacking concrete explainability features were excluded. Screening and data extraction followed PRISMA guidelines. Results: A total of 28 studies met the inclusion criteria. Ultrasound examinations (n = 9) and CT (n = 9) were the most common imaging modalities, followed by MRI (n = 6) and chest X-rays (n = 4). Clinical applications included disease classification (e.g., coronary artery disease and valvular heart disease) and the detection of myocardial or congenital abnormalities. Grad-CAM was the most frequently employed XAI method, followed by SHAP. Most studies used saliency-based techniques to generate visual explanations of model predictions. Conclusions: XAI holds considerable promise for improving the transparency and clinical acceptance of deep learning models in cardiovascular imaging. However, the evaluation of XAI methods remains largely qualitative, and standardization is lacking. Future research should focus on the robust, quantitative assessment of explainability, prospective clinical validation, and the development of more advanced XAI techniques beyond saliency-based methods. Strengthening the interpretability of AI models will be crucial to ensuring their safe, ethical, and effective integration into cardiovascular care.

Multimodal large language models (LLMs) that process text and images are increasingly discussed for medical imaging, yet their diagnostic performance on radiographs remains poorly characterized. We evaluated three commercially available multimodal LLMs (GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro) in a strict zero-shot setting for pediatric wrist fracture detection and compared them with a domain-trained Inception v3 convolutional neural network (CNN) on the same GRAZPEDWRI-DX dataset. We constructed a balanced patient-level test cohort of 1,000 children (2,298 radiographs; 500 fracture, 500 non-fracture). The CNN achieved high diagnostic performance (AUROC 0.905, AUPRC 0.920), whereas all LLMs performed close to chance (accuracies < 0.55, Matthews correlation coefficients ≈ 0) and produced bounding boxes often inconsistent with expert annotations. These findings indicate that, in a strict zero-shot setting, the three commercial multimodal LLMs evaluated here lack reliable diagnostic ability for pediatric wrist fracture detection and should therefore be regarded as exploratory research tools rather than clinically dependable systems for pediatric radiograph interpretation.

Background/Objectives: Lead acrylic shields (LASs) enhance radiation protection, with variation of their usage among different specialties. This prospective multicentric observational study aimed to explore patterns of ocular radiation exposure and LAS usage among interventionalists from different disciplines and to assess behavioral dose changes in a longitudinally monitored subgroup. Methods: From July 2019 to July 2020, ocular doses (protected and unprotected) were measured in 15 interventionalists from four specialties (radiology, neuroradiology, cardiology and vascular surgery) across 2286 procedures using thermoluminescent dosimeters. Procedure type, operator position, total body dose, fluoroscopy time, dose area product, LAS usage and dose exceedances were documented. Ocular doses of five interventionalists, who participated in our previous study, allowed a longitudinal comparison of radiation doses and protective behavior. Results: Cumulative annual unprotected ocular doses ranged from 0 to 59 mSv, with five participants exceeding the annual limit of 20 mSv, whereas protected doses (0 to 18 mSv) remained within the limits. Procedural LAS usage was seen in 89% of radiologists and 82% of vascular surgeons. Participants with 100% LAS usage recorded minimal measurable ocular doses, whereas lower LAS compliance resulted in higher ocular doses. Longitudinal subgroup data analyses showed a decrease in ocular dose per procedure in all five participants, suggesting a potential learning effect, without statistical significance. Conclusions: Unprotected ocular doses exceeded the annual occupational eye lens dose limit in several participants, whereas protected doses remained within recommended limits in all participants. No specialty demonstrated universally lower ocular doses, while consistent LAS use proved highly effective in reducing the dose in all groups. Longitudinal findings suggest that structured dose monitoring and repeated training may support behavioral improvements in radiation protection over time.

Background/Objectives: To characterize the zero-shot diagnostic behavior of three commercial multimodal large language models (MLLMs) on cardiopulmonary chest radiograph findings and to assess sex-stratified performance differences. Methods: GPT-5.4, Claude Opus 4.5, and Gemini 2.5 Pro were evaluated in 4500 pathology-specific radiograph evaluations based on frontal chest radiographs from the publicly available CheXpert dataset. Three balanced cohorts of 1500 images each were constructed for cardiomegaly, pulmonary edema, and pleural effusion (375 per sex-by-label subgroup). All models received identical zero-shot prompts requesting binary classification. The primary outcome was area under the receiver operating characteristic curve (AUC-ROC) with 95% bootstrap confidence intervals. Secondary outcomes were sensitivity and specificity. Results: A total of 4500 pathology-specific radiograph evaluations were performed across the three cohorts (2250 male and 2250 female cohort entries; mean age 58.4 ± 18.0 years). GPT-5.4 achieved the highest discrimination (AUC-ROC 0.836–0.883) but showed very low sensitivity (0.043–0.424) with near-perfect specificity (0.977–0.997). Claude Opus 4.5 showed moderate discrimination (AUC-ROC 0.698–0.761) with balanced sensitivity (0.396–0.876) and specificity (0.461–0.863). Gemini 2.5 Pro showed moderate discrimination (AUC-ROC 0.745–0.770) but favored sensitivity (0.673–0.973) at the expense of specificity (0.241–0.804). Sex-stratified analyses showed consistently higher AUC point estimates in male patients for cardiomegaly and pulmonary edema, but smaller and less directional differences for pleural effusion. Conclusions: Commercial MLLMs differ considerably in operating profiles, ranging from ultraconservative to aggressive detection, so that strong aggregate discrimination can mask sensitivity too low for reliable detection. None of the evaluated models are currently suitable for autonomous chest radiograph interpretation. Sex-stratified differences were modest but non-uniform, supporting subgroup-aware reporting rather than reliance on pooled metrics alone.

Objectives: To evaluate the prevalence, classification, and clinical relevance of incidental extracardiovascular findings in pre-transcatheter aortic valve implantation (TAVI) CT imaging. Methods: We conducted a retrospective single-center study of 225 patients undergoing pre-TAVI contrast-enhanced, ECG-gated CT scans between 2021 and 2023. Extracardiovascular findings were recorded and categorized into three groups based on presumed clinical relevance: Group A (findings with no need for follow-up), Group B (findings requiring follow-up), and Group C (findings requiring immediate intervention or treatment). Statistical analysis included a descriptive assessment of the overall prevalence of incidental findings and evaluation of age- and sex-related trends using chi-square tests with Bonferroni-adjusted pairwise comparisons. Results: The study cohort included 225 patients (53.3% male; mean age 79.9 ± 6.2 years, range 58–93). Extracardiovascular incidental findings were detected in 205 patients (91.1%). Among all 478 recorded findings, 82.6% were Group A, 14.4% Group B, and 2.9% Group C. On a per-patient level, 87.1% had at least one Group A finding, 24.9% had at least one Group B, and 6.2% had at least one Group C finding. Older age was associated with more incidental findings, with a significant difference observed between the 70–79 and 80–89 age groups (p = 0.002). No significant sex-related differences were found (p = 0.226). Findings were most frequently located in the abdomen (46.2%) and thorax (37.2%). Among all clinically relevant findings, the thorax was the most commonly affected region: 43.5% of Group B and 78.6% of Group C findings were located in the thorax, followed by the abdomen (33.3% of Group B and 7.1% of Group C findings). Conclusions: Extracardiovascular incidental findings are highly prevalent in pre-TAVI CT imaging and range from benign, age-related changes to potentially serious conditions such as malignancies or infections. Their presence reflects the comorbidity burden of the typical TAVI population and underscores the importance of recognizing non-vascular incidental findings in this clinical setting.

Pelvic kidney is a rare congenital anomaly characterized by aberrant location and complex vascular anatomy, predisposing it to complications such as trauma and hemorrhage. A 57-year-old man with a known right-sided pelvic kidney presented with acute macroscopic hematuria. Computed tomography revealed a large pericapsular hematoma with aberrant arterial supply and suspected active extravasation. A multidisciplinary approach was adopted, beginning with endovascular embolization as a preoperative measure to control bleeding, followed by transabdominal nephrectomy as definitive treatment. The combined approach ensured effective hemorrhage control and minimized surgical risk. This case highlights the importance of preoperative embolization as a safe and valuable adjunct in the management of hemorrhage in anatomically complex renal anomalies such as pelvic kidneys and underscores the role of multidisciplinary collaboration in achieving optimal outcomes.