Data Extraction from Free-Text Reports on Mechanical Thrombectomy in Acute Ischemic Stroke Using ChatGPT: A Retrospective Analysis.

Background Procedural details of mechanical thrombectomy in patients with ischemic stroke are important predictors of clinical outcome and are collected for prospective studies or national stroke registries. To date, these data are collected manually by human readers, a labor-intensive task that is prone to errors. Purpose To evaluate the use of the large language models (LLMs) GPT-4 and GPT-3.5 to extract data from neuroradiology reports on mechanical thrombectomy in patients with ischemic stroke. Materials and Methods This retrospective study included consecutive reports from patients with ischemic stroke who underwent mechanical thrombectomy between November 2022 and September 2023 at institution 1 and between September 2016 and December 2019 at institution 2. A set of 20 reports was used to optimize the prompt, and the ability of the LLMs to extract procedural data from the reports was compared using the McNemar test. Data manually extracted by an interventional neuroradiologist served as the reference standard. Results A total of 100 internal reports from 100 patients (mean age, 74.7 years ± 13.2 [SD]; 53 female) and 30 external reports from 30 patients (mean age, 72.7 years ± 13.5; 18 male) were included. All reports were successfully processed by GPT-4 and GPT-3.5. Of 2800 data entries, 2631 (94.0% [95% CI: 93.0, 94.8]; range per category, 61%-100%) data points were correctly extracted by GPT-4 without the need for further postprocessing. With 1788 of 2800 correct data entries, GPT-3.5 produced fewer correct data entries than did GPT-4 (63.9% [95% CI: 62.0, 65.6]; range per category, 14%-99%; P < .001). For the external reports, GPT-4 extracted 760 of 840 (90.5% [95% CI: 88.3, 92.4]) correct data entries, while GPT-3.5 extracted 539 of 840 (64.2% [95% CI: 60.8, 67.4]; P < .001). Conclusion Compared with GPT-3.5, GPT-4 more frequently extracted correct procedural data from free-text reports on mechanical thrombectomy performed in patients with ischemic stroke. © RSNA, 2024 Supplemental material is available for this article.

Impact of an AI software on the diagnostic performance and reading time for the detection of cerebral aneurysms on time of flight MR-angiography.

PURPOSE: To evaluate the impact of an AI-based software trained to detect cerebral aneurysms on TOF-MRA on the diagnostic performance and reading times across readers with varying experience levels. METHODS: One hundred eighty-six MRI studies were reviewed by six readers to detect cerebral aneurysms. Initially, readings were assisted by the CNN-based software mdbrain. After 6 weeks, a second reading was conducted without software assistance. The results were compared to the consensus reading of two neuroradiological specialists and sensitivity (lesion and patient level), specificity (patient level), and false positives per case were calculated for the group of all readers, for the subgroup of physicians, and for each individual reader. Also, reading times for each reader were measured. RESULTS: The dataset contained 54 aneurysms. The readers had no experience (three medical students), 2 years experience (resident in neuroradiology), 6 years experience (radiologist), and 12 years (neuroradiologist). Significant improvements of overall specificity and the overall number of false positives per case were observed in the reading with AI support. For the physicians, we found significant improvements of sensitivity on lesion and patient level and false positives per case. Four readers experienced reduced reading times with the software, while two encountered increased times. CONCLUSION: In the reading with the AI-based software, we observed significant improvements in terms of specificity and false positives per case for the group of all readers and significant improvements of sensitivity and false positives per case for the physicians. Further studies are needed to investigate the effects of the AI-based software in a prospective setting.

External evaluation of a deep learning-based approach for automated brain volumetry in patients with huntington's disease.

A crucial step in the clinical adaptation of an AI-based tool is an external, independent validation. The aim of this study was to investigate brain atrophy in patients with confirmed, progressed Huntington's disease using a certified software for automated volumetry and to compare the results with the manual measurement methods used in clinical practice as well as volume calculations of the caudate nuclei based on manual segmentations. Twenty-two patients were included retrospectively, consisting of eleven patients with Huntington's disease and caudate nucleus atrophy and an age- and sex-matched control group. To quantify caudate head atrophy, the frontal horn width to intercaudate distance ratio and the intercaudate distance to inner table width ratio were obtained. The software mdbrain was used for automated volumetry. Manually measured ratios and automatically measured volumes of the groups were compared using two-sample t-tests. Pearson correlation analyses were performed. The relative difference between automatically and manually determined volumes of the caudate nuclei was calculated. Both ratios were significantly different between the groups. The automatically and manually determined volumes of the caudate nuclei showed a high level of agreement with a mean relative discrepancy of - 2.3 ± 5.5%. The Huntington's disease group showed significantly lower volumes in a variety of supratentorial brain structures. The highest degree of atrophy was shown for the caudate nucleus, putamen, and pallidum (all p < .0001). The caudate nucleus volume and the ratios were found to be strongly correlated in both groups. In conclusion, in patients with progressed Huntington's disease, it was shown that the automatically determined caudate nucleus volume correlates strongly with measured ratios commonly used in clinical practice. Both methods allowed clear differentiation between groups in this collective. The software additionally allows radiologists to more objectively assess the involvement of a variety of brain structures that are less accessible to standard semiquantitative methods.

Clinical significance and prevalence of subarachnoid hyperdensities on flat detector CT after mechanical thrombectomy: does it really matter?

BACKGROUND: Subarachnoid hyperdensities after mechanical thrombectomy (MT) are a common finding. However, it is often regarded as clinically insignificant. OBJECTIVE: With this single-center investigation, to identify the prevalence of subarachnoid hyperdensities following MT, associated predictors, and the impact on the clinical outcome of the patients. METHODS: 383 patients from the stroke registry were analyzed for the presence of subarachnoid hyperdensities on flat detector CT (FDCT) directly after the completion of MT, and on follow-up dual-energy CT, then classified according to a visual grading scale. 178 patients were included with anterior circulation occlusions. Regression analysis was performed to identify significant predictors, and Kruskal-Wallis analysis and Χ2 test were performed to test the variables among the different groups. The primary outcome was the modified Rankin Scale (mRS) score at 90 days and was analyzed with the Wilcoxon-Mann-Whitney rank-sum test. RESULTS: The prevalence of subarachnoid hyperdensities on FDCT was (66/178, 37.1%) with patients experiencing a significant unfavorable outcome (P=0.035). Significantly fewer patients with subarachnoid hyperdensities achieved a mRS score of ≤3 at 90 days 25/66 (37.9%) vs 60/112 (53.6%), P=0.043). In addition, mortality was significantly higher in the subarachnoid hyperdensities group (34.8% vs 19.6%, P=0.024). Distal occlusions and a higher number of device passes were significantly associated with subarachnoid hyperdensities (P=0.026) and (P=0.001), respectively. Patients who received intravenous tissue plasminogen activator had significantly fewer subarachnoid hyperdensities (P=0.029). CONCLUSIONS: Postinterventional subarachnoid hyperdensities are a frequent finding after MT and are associated with neurological decline and worse functional outcome. They are more common with distal occlusions and multiple device passes.

Brain Volume Changes after COVID-19 Compared to Healthy Controls by Artificial Intelligence-Based MRI Volumetry.

Cohort studies that quantify volumetric brain data among individuals with different levels of COVID-19 severity are presently limited. It is still uncertain whether there exists a potential correlation between disease severity and the effects of COVID-19 on brain integrity. Our objective was to assess the potential impact of COVID-19 on measured brain volume in patients with asymptomatic/mild and severe disease after recovery from infection, compared with healthy controls, using artificial intelligence (AI)-based MRI volumetry. A total of 155 participants were prospectively enrolled in this IRB-approved analysis of three cohorts with a mild course of COVID-19 (n = 51, MILD), a severe hospitalised course (n = 48, SEV), and healthy controls (n = 56, CTL) all undergoing a standardised MRI protocol of the brain. Automated AI-based determination of various brain volumes in mL and calculation of normalised percentiles of brain volume was performed with mdbrain software, using a 3D T1-weighted magnetisation-prepared rapid gradient echo (MPRAGE) sequence. The automatically measured brain volumes and percentiles were analysed for differences between groups. The estimated influence of COVID-19 and demographic/clinical variables on brain volume was determined using multivariate analysis. There were statistically significant differences in measured brain volumes and percentiles of various brain regions among groups, even after the exclusion of patients undergoing intensive care, with significant volume reductions in COVID-19 patients, which increased with disease severity (SEV > MILD > CTL) and mainly affected the supratentorial grey matter, frontal and parietal lobes, and right thalamus. Severe COVID-19 infection, in addition to established demographic parameters such as age and sex, was a significant predictor of brain volume loss upon multivariate analysis. In conclusion, neocortical brain degeneration was detected in patients who had recovered from SARS-CoV-2 infection compared to healthy controls, worsening with greater initial COVID-19 severity and mainly affecting the fronto-parietal brain and right thalamus, regardless of ICU treatment. This suggests a direct link between COVID-19 infection and subsequent brain atrophy, which may have major implications for clinical management and future cognitive rehabilitation strategies.

School of Thrombectomy-A 3-Step Approach to Perform Acute Stroke Treatment with Simulator Training and Virtual Supervision by Remote Streaming Support (RESS).

As the number of neurointerventional procedures continues to increase, so does the need for well-trained neurointerventionalists. The purpose of this work was to establish and assess a systematic 3­step approach to perform acute stroke treatment including simulator training and virtual supervision by remote streaming support (RESS). Five trainees (four men, one women) who have completed the 3­step approach have answered an 11-item questionnaire (5-point Likert scale) in order to evaluate training step 1 (simulator). Furthermore, all trainees and one supervisor (female) answered a standardized questionnaire following the initial 15 consecutive thrombectomies for each trainee, corresponding to a total of 75 thrombectomies. The simulator training yielded learning benefits and confidence gain to perform MT on patients. The RESS approach facilitated the translation during the first independently performed thrombectomies on patients. In summary, the presented 3­step approach increases the level of safety, as reported by the trainees and supervisor in this study and may enable an accelerated training of neurointerventionalists.

Comparative assessment of woven endobridge embolization and standard coil occlusion for the treatment of ruptured basilar tip aneurysms.

PURPOSE: Endovascular coil occlusion represents the standard treatment for basilar tip aneurysms. Recently, this role has been rivalled by intrasaccular flow disruptors across numerous centres. We retrospectively compared WEB embolization and coiling for the treatment of ruptured basilar tip aneurysms. METHODS: Patients treated with WEB or coiling at four neurovascular centres were reviewed. Procedure-related complications, clinical outcome, and angiographic results were retrospectively compared. RESULTS: The study included 23 patients treated with the WEB (aneurysm size: 6.6 ± 1.9 mm) and 56 by coiling (aneurysm size: 6.7 ± 2.5 mm). Stent-assistance was more often necessary with coiling than with WEB embolization (32% vs. 4%, p = 0.009). A modified Rankin scale score ≤ 2 at discharge had 21 (37.5%) patients in the coiling group and 12 (52.2%) in the WEB group (p = 0.235). Immediate complete and adequate occlusion rates were 52% for the WEB and 87% for coiling. At short-term follow-up, these rates were 87% for the WEB and 72% for coiling, respectively. There was no delayed aneurysm re-bleeding during follow-up. CONCLUSION: Both coiling and WEB seem to prevent rebleeding in ruptured BTA aneurysms. WEB embolization required less frequently stent-support than coiling, potentially advantageous for SAH patients to avoid anti-platelet therapy in the light of concomitant procedures like ventricular drainage.

First Experience with the Nimbus Stentretriever : A Novel Device to Handle Fibrin-rich Clots.

PURPOSE: To share our first experience with the Nimbus stentretriever, a multizone device designed to assist neurointerventionalists in handling fibrin-rich clots in endovascular stroke treatment. METHODS: We retrospectively analyzed the data of patients who were treated with the Nimbus stentretriever at our high-volume stroke center between May 2021 and May 2022. We evaluated the number of passes before Nimbus was used, the number of passes with nimbus, as well as the recanalization success before and after Nimbus according to the modified treatment in cerebral ischemia (mTICI) scale. Also, patient characteristics, procedural times and clinical outcomes were documented. RESULTS: A total of 21 consecutive patients were included in the study. An mTICI 2b/3 could be achieved in 76.2% and mTICI 2c/3 could be achieved in 57.1%. The mean number of passes was 3.4 before the use of Nimbus, 2.2 with Nimbus, and 5.4 for all passes with and without Nimbus and 4 occlusions (19.0%) were successfully recanalized with direct aspiration after the use of Nimbus. We observed seven subarachnoid hemorrhages (33.3%) and two cases of vasospasm. CONCLUSION: In our series, the use of Nimbus resulted in successful recanalization in half of the patients after otherwise unsuccessful thrombectomy maneuvers; therefore, it should be considered as a rescue option if the maneuver with conventional stent retrievers was unsuccessful.

Comparison of angiographic outcomes and complication rates of WEB embolization and coiling for treatment of unruptured basilar tip aneurysms.

Endovascular coiling represents the standard treatment for basilar tip aneurysms. Some of these aneurysms are not amenable to conventional coiling due to a complex aneurysm geometry, hence, novel devices such as the Woven Endobridge (WEB) have been developed. We retrospectively compared WEB embolization and coiling for the treatment of unruptured basilar tip aneurysms. Patients treated with WEB or coiling at four centers were reviewed. Procedure-related complications, clinical outcome and angiographic results were retrospectively evaluated and compared. Forty patients treated with the WEB and 35 patients treated by coiling were included. Stent-assistance was more often necessary for coiling than for WEB embolization (71% vs 2.5%, p < 0.001). The technical success rates were 100% for both methods. The overall complication rates were not significantly different between groups (WEB: 5%, coil: 11%, p = 0.409). Procedural morbidity rates were 9% in the coiling group and 2.5% in the WEB group (p = 0.334). There was no mortality. Treatment duration was shorter for WEB implantation than for coiling (p = 0.048). At mid-term follow-up, complete occlusion, neck remnants and aneurysm remnants were observed in 89%, 4% and 7% for the WEB, respectively, and in 100%, 0% and 0% for coiling. While complication rates and mid-term angiographic outcome was comparable between the groups, the WEB was associated with a shorter treatment duration and required stent-assistance less frequently. The choice of the treatment modality should be made based on the specific aneurysm characteristics, the individual experience of the neurointerventionalist and patient preference.

Late-Onset Vaccine-Induced Immune Thombotic Thrombocytopenia (VITT) with Cerebral Venous Sinus Thrombosis.

OBJECTIVES: Vaccine-induced thrombotic thrombocytopenia (VITT) is a rare complication after adenoviral vector vaccination against COVID-19 reported up to 24 days after ChAdOx1 nCOV-19 (AZD1222) vaccination. This report describes a case with a significantly later onset of VITT with cerebral venous sinus thrombosis. CASE DESCRIPTION: We report a 42-year-old woman presenting to the emergency department 53 days after AZD1222 vaccination with sudden onset sensory aphasia and an 18-day history of headache. Cranial computed tomography (CT) showed acute intracranial hemorrhage and CT venogram demonstrated thrombosis of the left vein of Labbé and transverse and sigmoid sinus. D-dimers were elevated and despite a normal platelet count, platelet-activating anti-PF4 antibody testing was positive, confirming the diagnosis of VITT. The patient was treated with intravenous immunoglobulins and argatroban, and was discharged without any neurological deficit on day 12. CONCLUSION: Our report of VITT with symptom onset on day 35 and diagnosis of cerebral sinuous thrombosis on day 53 after AZD1222 vaccination significantly enhances the time window during which VITT may occur.

Detection of Degenerative Changes on MR Images of the Lumbar Spine with a Convolutional Neural Network: A Feasibility Study.

Our objective was to evaluate the diagnostic performance of a convolutional neural network (CNN) trained on multiple MR imaging features of the lumbar spine, to detect a variety of different degenerative changes of the lumbar spine. One hundred and forty-six consecutive patients underwent routine clinical MRI of the lumbar spine including T2-weighted imaging and were retrospectively analyzed using a CNN for detection and labeling of vertebrae, disc segments, as well as presence of disc herniation, disc bulging, spinal canal stenosis, nerve root compression, and spondylolisthesis. The assessment of a radiologist served as the diagnostic reference standard. We assessed the CNN's diagnostic accuracy and consistency using confusion matrices and McNemar's test. In our data, 77 disc herniations (thereof 46 further classified as extrusions), 133 disc bulgings, 35 spinal canal stenoses, 59 nerve root compressions, and 20 segments with spondylolisthesis were present in a total of 888 lumbar spine segments. The CNN yielded a perfect accuracy score for intervertebral disc detection and labeling (100%), and moderate to high diagnostic accuracy for the detection of disc herniations (87%; 95% CI: 0.84, 0.89), extrusions (86%; 95% CI: 0.84, 0.89), bulgings (76%; 95% CI: 0.73, 0.78), spinal canal stenoses (98%; 95% CI: 0.97, 0.99), nerve root compressions (91%; 95% CI: 0.89, 0.92), and spondylolisthesis (87.61%; 95% CI: 85.26, 89.21), respectively. Our data suggest that automatic diagnosis of multiple different degenerative changes of the lumbar spine is feasible using a single comprehensive CNN. The CNN provides high diagnostic accuracy for intervertebral disc labeling and detection of clinically relevant degenerative changes such as spinal canal stenosis and disc extrusion of the lumbar spine.

Diagnostic Accuracy of Quantitative Imaging Biomarkers in the Differentiation of Benign and Malignant Vertebral Lesions : Combination of Diffusion-Weighted and Proton Density Fat Fraction Spine MRI.

PURPOSE: To compare and combine the diagnostic performance of the apparent diffusion coefficient (ADC) derived from diffusion-weighted imaging (DWI) and proton density fat fraction (PDFF) derived from chemical-shift encoding (CSE)-based water-fat magnetic resonance imaging (MRI) for distinguishing benign and malignant vertebral bone marrow lesions (VBML). METHODS: A total of 55 consecutive patients with 53 benign (traumatic, inflammatory and primary) and 36 malignant (metastatic and hematologic) previously untreated VBMLs were prospectively enrolled in this IRB-approved study and underwent sagittal DWI (single-shot spin-echo echo-planar with multi-slice short TI inversion recovery fat suppression) and CSE-based MRI (gradient-echo 6­point modified Dixon) in addition to routine clinical spine MRI at 1.5 T or 3.0 T. Diagnostic reference standard was established according to histopathology or imaging follow-up. The ADC = ADC (0, 800) and PDFF = fat / (water + fat) were calculated voxel-wise and examined for differences between benign and malignant lesions. RESULTS: The ADC and PDFF values of malignant lesions were significantly lower compared to benign lesions (mean ADC 861â€¯× 10-6 mm2/s vs. 1323â€¯× 10-6 mm2/s, p < 0.001; mean PDFF 3.1% vs. 28.2%, p < 0.001). The areas under the curve (AUC) and diagnostic accuracies were 0.847 (p < 0.001) and 85.4% (cut-off at 1084.4â€¯× 10-6 mm2/s) for ADC and 0.940 (p < 0.001) and 89.9% for PDFF (cut-off at 7.8%), respectively. The combined use of ADC and PDFF improved the diagnostic accuracy to 96.6% (malignancy if ADC ≤ 1118.2â€¯× 10-6 mm2/s and PDFF ≤ 20.0%, otherwise benign). CONCLUSION: Quantitative evaluation of both ADC and PDFF was useful in differentiating benign VBMLs from malignancy. The combination of ADC and PDFF improved the diagnostic performance and yielded high diagnostic accuracy for the differentiation of benign and malignant VBMLs.

Proton Density Fat Fraction Spine MRI for Differentiation of Erosive Vertebral Endplate Degeneration and Infectious Spondylitis.

Vertebral Modic type 1 (MT1) degeneration may mimic infectious disease on conventional spine magnetic resonance imaging (MRI), potentially leading to additional costly and invasive investigations. This study evaluated the diagnostic performance of the proton density fat fraction (PDFF) for distinguishing MT1 degenerative endplate changes from infectious spondylitis. A total of 31 and 22 patients with equivocal diagnosis of MT1 degeneration and infectious spondylitis, respectively, were retrospectively enrolled in this IRB-approved retrospective study and examined with a chemical-shift encoding (CSE)-based water-fat 3D six-echo modified Dixon sequence in addition to routine clinical spine MRI. Diagnostic reference standard was established according to histopathology or clinical and imaging follow-up. Intravertebral PDFF [%] and PDFFratio (i.e., vertebral endplate PDFF/normal vertebrae PDFF) were calculated voxel-wise within the single most prominent edematous bone marrow lesion per patient and examined for differences between MT1 degeneration and infectious spondylitis. Mean PDFF and PDFFratio of infectious spondylitis were significantly lower compared to MT1 degenerative changes (mean PDFF, 4.28 ± 3.12% vs. 35.29 ± 17.15% [p < 0.001]; PDFFratio, 0.09 ± 0.06 vs. 0.67 ± 0.37 [p < 0.001]). The areas under the curve (AUC) and diagnostic accuracies were 0.977 (p < 0.001) and 98.1% (cut-off at 12.9%) for PDFF and 0.971 (p < 0.001) and 98.1% (cut-off at 0.27) for PDFFratio. Our data suggest that quantitative evaluation of vertebral PDFF can provide a high diagnostic accuracy for differentiating erosive MT1 endplate changes from infectious spondylitis.

Fibroblast growth factor receptor 1 gene amplification in gastric adenocarcinoma.

Gastric adenocarcinomas are associated with a poor prognosis due to the fact that the tumor has often metastasized by the time of diagnosis. Thus, identification of novel therapeutic targets is highly desirable. Here, we examined gene copy number of fibroblast growth factor receptor 1 (FGFR1), a potential target for tyrosine kinase inhibitors, and clinicopathologic parameters in a large cohort of gastric adenocarcinomas. We performed fluorescence in situ hybridization analysis of 293 gastric adenocarcinomas using tissue microarrays. Amplification of the FGFR1 gene is a rare but noticeable event that can be found in 2% (6/293) of cases and was associated with poor 10-year survival (median 15.3 months in FGFR1-amplified cases versus 36 months in nonamplified cases, P = .047) and a higher rate of distant metastasis (P = .025). FGFR1 appears to represent a potential new therapeutic target in a subset of patients with gastric carcinoma. Identification of gastric cancers harboring FGFR1 amplification may be important in preselecting patients and/or interpreting clinical studies using tyrosine kinase inhibitors.

Fibroblast growth factor receptor 1 gene amplification in pancreatic ductal adenocarcinoma.

AIMS: Pancreatic ductal adenocarcinomas (PDACs) are chemoresistant, resulting in extremely poor survival of patients; therefore, novel molecular targets, even in small subsets of genetically characterized tumours, are urgently needed. Tyrosine kinase receptor inhibitors (TKIs) are already in clinical use. The aims of this study were to examine the gene copy number and expression of fibroblast growth factor receptor 1 (FGFR1) in 155 patients with PDAC, and investigate the effects of the FGFR-specific inhibitor BGJ398 on FGFR1-amplified pancreatic tumour cells in vitro. METHODS AND RESULTS: Fluorescence in-situ hybridization (FISH) and immunohistochemical analysis of 155 PDACs were performed using tissue microarrays. Amplification of FGFR1 was found in 2.6% (4/155) of cases. Four per cent of tumours (5/125) were shown to express FGFR1 by immunohistochemistry. Sequence analysis demonstrated an activating KRAS mutation (exon 2) in all FGFR1-amplified cases. The FGFR1-amplified pancreatic carcinoma cell line PT45P1 showed high levels of FGFR1 mRNA and protein expression. Proliferation of this cell line can be inhibited using the FGFR1 inhibitor BGJ398. CONCLUSIONS: FGFR1 represents a potential new therapeutic target in a subset of patients harbouring FGFR1-amplified tumours. Identification of pancreatic cancers harbouring FGFR1 amplification may be important in preselecting patients and/or interpreting clinical studies using TKIs.