Lukas Goertz

Background: Symptom-based localization of brainstem ischemia is challenging because of the anatomical complexity of the brainstem and the nonspecific overlap of clinical syndromes. Whether large language models (LLMs) can meaningfully assist in this task remains uncertain. Objective: This study aimed to compare the performance of several OpenAI LLMs and neurologists in localizing diffusion-weighted imaging (DWI)–confirmed brainstem ischemic lesions based on symptom descriptions alone. Methods: In this retrospective single-center study, 109 patients with DWI-confirmed acute brainstem ischemia were included. Three neurologists and 6 LLMs (GPT-5, GPT-4, GPT-4.1, GPT-4o, o3, and o3-pro) predicted lesion localization using a combined anatomical-lateral end point (left or right midbrain, pons, and medulla) based on symptom descriptions alone. Overall and regional accuracy, the Cohen κ, 6-class confusion matrices, and point-biserial correlations between symptom count and correct prediction were assessed. Because all raters evaluated the same cases, paired McNemar tests with Benjamini-Hochberg correction were used for pairwise performance comparisons. Results: GPT-4 and GPT-4o achieved the highest overall accuracy (61/109, 56%; 95% CI 46.1%‐65.5%). Agreement with the DWI reference standard remained limited across all raters, with the Cohen κ reaching a maximum of 0.291 for GPT-4o. Confusion matrices showed that higher performance was driven mainly by pontine cases, whereas misclassification remained frequent in mesencephalic and medullary lesions. Regional analyses outside the pons were imprecise because mesencephalic and medullary subgroups each contained only 16 cases. A higher number of documented symptoms was associated with correct prediction for GPT-4, GPT-5, GPT-o3, and 1 neurologist. Conclusions: Although some LLMs showed higher relative accuracy than the participating neurologists, absolute performance remained limited and clinically insufficient. These findings are best interpreted as an exploratory benchmark under constrained conditions: absolute performance remained modest, agreement beyond chance was limited, and performance outside pontine lesions was inconclusive.

Background/Objectives: Accurate and reproducible lymph node (LN) measurement is essential for oncologic staging and therapy monitoring but is subject to inter-reader variability. This study evaluated the accuracy and reproducibility of a fully automated artificial intelligence (AI)-based LN measurement workflow in contrast-enhanced chest CT, using multi-reader manual measurements as reference. Methods: Sixty thoracic LNs from seven patients were independently measured by 13 radiologists in two reading rounds. The median of all measurements served as the ground truth (GT). Automated short- and long-axis diameters were derived from fully automated 3D CNN-based segmentations. Agreement between AI and manual measurements was assessed using Friedman testing, intraclass correlation coefficients (ICCs), and concordance correlation coefficients (CCCs). Measurement stability was evaluated across repeated runs on different hardware systems. Results: A total of 2280 manual measurements were analyzed. Manual assessment showed significant inter-reader variability (p < 0.01), while intra-reader agreement was high. No significant differences were observed between AI-based measurements and the GT (all p > 0.01). Agreement was good, with CCC values of 0.86 (SAD) and 0.79 (LAD). AI-based measurements were numerically stable across hardware configurations. Conclusions: Fully automated AI-based LN measurements in chest CT scans provide strong agreement with multi-reader consensus and high numerical stability. Automated measurement may support more standardized and reproducible oncologic imaging assessment.

Background/Objectives: To evaluate the diagnostic accuracy of two visual large language models (vLLMs), GPT-4o (OpenAI) and Claude Sonnet 3.5 (Anthropic), for detecting brain metastases in routine MRI using combined imaging and textual input. Methods: This retrospective study included 31 patients with and 46 without brain metastases with underlying melanoma (n = 24), lung cancer (n = 23), breast cancer (n = 17), or renal cell carcinoma (n = 13). In total, 100 MRI examinations (50 with, 50 without metastases) were provided to both vLLMs using a single representative slice per sequence, together with clinical history and the referring question. The generated free-text reports were evaluated for detection accuracy, overdiagnosis, correct sequence recognition, anatomical localization, lesion laterality, and lesion size estimation. Results: Both vLLMs showed perfect sensitivity (100% for both) but very low specificity (GPT-4o: 8%, Sonnet 3.5: 4%; p = 0.625), resulting in low diagnostic accuracy (GPT-4o: 54%, Sonnet 3.5: 52%; p = 0.625). Sequence identification was highly accurate in both models, with GPT-4o performing significantly better (100% vs. 93%; p < 0.05). Identification of the anatomical brain region (70% vs. 72%; p = 1.00) and lesion laterality (62% vs. 76%; p = 0.189) was comparable. Both models hallucinated additional lesions in 12% of cases. Lesion size measurements showed no significant differences between the models or in comparison with the radiologist. Conclusions: GPT-4o and Claude Sonnet 3.5 can generate radiological reports and detect brain metastases with excellent sensitivity, but their very low specificity, frequent hallucinations, and limited spatial reliability currently preclude clinical application. Future work should address how the balance between visual and textual input influences diagnostic behavior in vLLMs.

Purpose Surface modifications may improve procedural safety of flow diverter (FD) treatment, but larger studies are rare. This multicenter study aims to evaluate the safety and efficacy of unruptured aneurysm treatment with surface-modified and coated FDs. Methods Patients treated with the following FDs at 8 neurovascular centers were retrospectively reviewed: Pipeline Vantage Embolization Device, Pipeline Flex with Shield Technology, FRED X, p48/64 HPC, Derivo Embolization Device (DED), and DED 2heal. Aneurysm characteristics, procedural details, and angiographic results were evaluated in detail. Results A total of 511 patients with 545 aneurysms were included (mean aneurysm size: 8.2 ± 5.5 mm, 12% posterior circulation, 14% non-saccular morphology). Multiple FDs were used in 26/515 (5%) procedures and adjunctive coiling in 76 (15%). Thromboembolic events occurred in 40 (7.8%) cases, including 9 (1.7%) major ischemic events, 20 (3.9%) minor ischemic events and 11 (2.1%) asymptomatic/technical events. Non-saccular aneurysm morphology (p = 0.014) and use of multiple FDs (p = 0.025) favored thromboembolic events. Hemorrhagic events occurred in 5 (1.0%) cases, of which 3 were symptomatic. The combined morbidity and mortality rate was 11/515 (2.1%). Complete and adequate occlusion rates were 66% (196/297) and 80% (238/297) at 6 months, respectively, and 70% (64/92) and 85% (78/92) at 12 months, respectively. Conclusion The present study demonstrates high safety and efficacy for coated or surface-modified FDs. Whether the surface modifications confer a clinical benefit needs to be addressed in comparative studies.

Purpose: Small, shallow aneurysms (SSAs) present technical challenges for endovascular treatment with the risk of aneurysm perforation and coil protrusion. The Woven EndoBridge (WEB) device, particularly the lowest height of 2 mm variant, offers a potential alternative option for these aneurysms. This multicenter study evaluates the feasibility, safety and efficacy of the WEB for SSAs ≤ 4.1 mm width. Methods: A total of 103 aneurysms (mean size: 3.4 ± 1.1 mm, 67 [65%] bifurcation) treated with WEB single-layer sizes ranging from 3 × 2 mm to 4.5 × 2 mm were retrospectively analyzed. Data on procedural success, complications, and angiographic outcomes were collected. The incidence and potential strategies to counteract WEB protrusion were evaluated. Results: WEB deployment was successful in 97 (94%) cases, but WEB protrusion occurred in 12 (12%) cases. Strategies to counteract WEB protrusion were change to another endovascular technique ( n = 4), additional stent implantation (n = 4), change to a smaller WEB size ( n = 3), and repositioning of the WEB ( n = 1). Procedural thromboembolic complications occurred in 4 (4%) cases, of which 3 were due to protrusion of the WEB, but all were asymptomatic. Two hemorrhagic complications (2%), caused by aneurysm perforation with the microwire in the first case and with the folded WEB tip in the second, resulted in procedural morbidity (2%). Complete and adequate occlusion rates were 82% (42/51) and 90% (46/51) at 6 months and 76% (19/25) and 88% (22/25) at 12 months, respectively. Conclusion: The 2 mm height WEB demonstrated feasibility and efficacy in the treatment of SSAs with good safety results, making it a viable alternative endovascular option for SSAs.

Introduction: Acute appendicitis (AA) is a common cause of abdominal pain that can lead to complications like perforation and intra-abdominal abscesses, increasing morbidity and mortality, often requiring emergency surgery. Nevertheless, appendectomy is performed in up to 95% of uncomplicated cases, while complications like perforation and intra-abdominal abscesses increase morbidity and mortality. The current study compares the accuracy of GPT-4.5, DeepSeek R1, and machine learning in assisting with surgical decision-making for patients presenting with lower abdominal pain at the Emergency Department. Methods: In this multicenter retrospective study, 63 histopathologically confirmed appendicitis patients and 50 control patients with right abdominal pain presenting at the Emergency Department at two German hospitals between October 2022 and October 2023 were included. Using each patient’s clinical, laboratory, and radiological findings, DeepSeek (with and without Retrieval-Augmented Generation using 2020 Jerusalem guidelines) was compared in terms of accuracy with GPT-4.5 and a random forest-based machine-learning model, with a board-certified surgeon (reference standard) to determine the optimal treatment approach (laparoscopic exploration/appendectomy versus conservative antibiotic therapy). Results: Accuracy of agreement with board-certified surgeons in the decision-making of appendectomy versus conservative therapy increased non-significantly from 80.5% to 83.2% with DeepSeek and from 70.8 to 76.1% when GPT-4.5 was provided with the World Journal of Emergency Surgery 2020 Jerusalem guidelines on the diagnosis and treatment of acute appendicitis. The estimated machine-learning model training accuracy was 84.3%, while the validation accuracy for the model was 85.0%. Discussion: GPT-4.5 and DeepSeek R1, as well as the machine-learning model, demonstrate promise in aiding surgical decision-making for appendicitis, particularly in resource-constrained settings. Ongoing training and validation are required to optimize the performance of such models.

Purpose: Recent advancements in flow-diverter (FD) technology have introduced various coating strategies to reduce platelet aggregation and thromboembolic complications in intracranial aneurysm treatment. This study compared postinterventional diffusion-weighted imaging (DWI) lesions, as markers of thromboembolic events, between coated and uncoated FDs. Methods: A retrospective analysis was conducted on patients treated for intracranial aneurysms between June 2018 and December 2024. The primary outcome was the occurrence of postinterventional DWI + lesions, with periprocedural factors analyzed for their association. Results: A total of 91 patients with 104 aneurysms were treated with 43 uncoated ( n = 30 FRED, n = 13 DERIVO) and 51 coated FDs ( n = 36 FREDX, n = 15 DERIVO2Heal). DWI + lesions were observed in 39.5% of uncoated and 54.9% of coated FD cases ( p = 0.1). After 1:1 propensity score matching, no significant difference was found ( p = 0.131). Contrast stasis ( p = 0.025) within aneurysm sac after FD deployment, FD proximal/distal ratio ( p = 0.007), FD size ( p = 0.047), and intervention duration ( p = 0.036) were significantly associated with DWI + lesions. In multivariable analysis, contrast stasis (ORa = 3.7, 95% CI 1.1–11.9, p = 0.029), FD proximal/distal ratio (ORa = 27.8, 95% CI 1.8–426.8, p = 0.017), and older age (ORa = 1.06, 95% CI 1.00–1.11, p = 0.050) increased incidence of DWI + lesions. Conclusion: In this single-center observational study, coated FRED and DERIVO FDs did not reduce postinterventional DWI + lesions compared to their uncoated counterparts.

Objective: Phenox flow diverters (p64 and its smaller vessel variant p48) represent an established treatment option for intracranial aneurysms. This study evaluates the safety and efficacy of the new generation of these devices with an additional antithrombotic surface coating (HPC). Methods: Consecutive patients treated between 2020 and 2023 at three institutions were retrospectively reviewed for aneurysm characteristics, procedural details, complications, and angiographic outcomes. Results: Sixty-one patients (mean age 56 years) were treated for 61 aneurysms. The mean aneurysm size was 8.3 ± 4.9 mm, 12 (19.7%) were ruptured, 16 (26.2%) were recurrent after previous treatment, 10 (16.4%) were located in the posterior circulation, and 7 (11.5%) had nonsaccular morphology. All procedures were technically successful, with a single device sufficient in 60/61 (98%) cases. Delivery problems included device twisting in one case and incomplete proximal opening in another. Additional angioplasty was performed in 3/61 (4.9%) procedures and additional coiling in 4 (6.6%). There were 3 (4.9%) major events (1 thromboembolic ischemic stroke, 1 fatal intracranial hemorrhage, and 1 delayed aneurysm rupture) and 6 (9.8%) minor strokes. Follow up at a mean of 6 months showed complete occlusion in 33/42 (79%) aneurysms, neck remnants in 5 (12%), and aneurysm remnants in 4 (10%). Conclusions: The results demonstrate comparable acceptable complication rates and angiographic results of the Phenox HPC, which are similar to other flow diverters. Long-term and comparative studies are needed to evaluate the full potential of these devices.

Background: The latest advancement of artificial intelligence (AI) is generative pretrained transformer large language models (LLMs). They have been trained on massive amounts of text, enabling humanlike and semantical responses to text-based inputs and requests. Foreshadowing numerous possible applications in various fields, the potential of such tools for medical data integration and clinical decision-making is not yet clear. Objective: In this study, we investigate the potential of LLMs in report-based medical decision-making on the example of acute ischemic stroke (AIS), where clinical and image-based information may indicate an immediate need for mechanical thrombectomy (MT). The purpose was to elucidate the feasibility of integrating radiology report data and other clinical information in the context of therapy decision-making using LLMs. Methods: A hundred patients with AIS were retrospectively included, for which 50% (50/100) was indicated for MT, whereas the other 50% (50/100) was not. The LLM was provided with the computed tomography report, information on neurological symptoms and onset, and patients’ age. The performance of the AI decision-making model was compared with an expert consensus regarding the binary determination of MT indication, for which sensitivity, specificity, and accuracy were calculated. Results: The AI model had an overall accuracy of 88%, with a specificity of 96% and a sensitivity of 80%. The area under the curve for the report-based MT decision was 0.92. Conclusions: The LLM achieved promising accuracy in determining the eligibility of patients with AIS for MT based on radiology reports and clinical information. Our results underscore the potential of LLMs for radiological and medical data integration. This investigation should serve as a stimulus for further clinical applications of LLMs, in which this AI should be used as an augmented supporting system for human decision-making.