Purpose: Recently, a large number of patients with acute ischemic stroke benefited from the use of thrombectomy, a minimally invasive intervention technique for mechanically removing thrombi from the cerebrovasculature. During thrombectomy, 2D digital subtraction angiography (DSA) image sequences are acquired simultaneously from the posterior-anterior and the lateral view to control whether thrombus removal was successful, and to possibly detect newly occluded areas caused by thrombus fragments split from the main thrombus. However, such new occlusions, which would be treatable by thrombectomy, may be overlooked during the intervention. To prevent this, we developed a deep learning-based approach to automatic classification of DSA sequences into thrombus-free and non-thrombus-free sequences. Methods: We performed a retrospective study based on the single-center DSA data of thrombectomy patients. For classifying the DSA sequences, we applied Long Short-Term Memory or Gated Recurrent Unit networks and combined them with different Convolutional Neural Networks used as feature extractor. These network variants were trained on the DSA data by using five-fold cross-validation. The classification performance was determined on a test data set with respect to the Matthews correlation coefficient (MCC) and the area under the curve (AUC). Finally, we evaluated our models on patient cases, in which overlooking thrombi during thrombectomy had happened. Results: Depending on the specific model configuration used, we obtained a performance of up to 0.77 $$\mid $$∣ 0.94 for the MCC $$\mid $$∣ AUC, respectively. Additionally, overlooking thrombi could have been prevented in the reported patient cases, as our models would have classified the corresponding DSA sequences correctly. Conclusion: Our deep learning-based approach to thrombus identification in DSA sequences yielded high accuracy on our single-center test data set. External validation is now required to investigate the generalizability of our method. As demonstrated, using this new approach may help reduce the incident risk of overlooking thrombi during thrombectomy in the future.

Cone beam computed tomography (CBCT) is increasingly used for dental and maxillofacial imaging. The occurrence of incidental findings has been reported, but clinical implications of these findings remain unclear. The study’s aim was to identify the frequency and clinical impact of incidental findings in CBCT. A total of 374 consecutive CBCT examinations of a 3 year period were retrospectively evaluated for the presence, kind, and clinical relevance of incidental findings. In a subgroup of 54 patients, therapeutic consequences of CBCT incidental findings were queried from the referring physicians. A total of 974 incidental findings were detected, involving 78.6% of all CBCT, hence 2.6 incidental findings per CBCT. Of these, 38.6% were classified to require treatment, with an additional 25.2% requiring follow-up. Incidental findings included dental pathologies in 55.3%, pathologies of the paranasal sinuses and airways in 29.2%, osseous pathologies in 14.9% of all CBCT, and findings in the soft tissue or TMJ in few cases. Clinically relevant dental incidental findings were detected significantly more frequently in CBCT for implant planning compared to other indications (60.7% vs. 43.2%, p < 0.01), and in CBCT with an FOV ≥ 100 mm compared to an FOV < 100 mm (54.7% vs. 40.0%, p < 0.01). Similar results were obtained for paranasal incidental findings. In a subgroup analysis, 29 of 54 patients showed incidental findings which were previously unknown, and the findings changed therapeutical management in 19 patients (35%). The results of our study highlighted the importance of a meticulous analysis of the entire FOV of CBCT for incidental findings, which showed clinical relevance in more than one in three patients. Due to a high number of clinically relevant incidental findings especially in CBCT for implant planning, an FOV of 100 × 100 mm covering both the mandible and the maxilla was concluded to be recommendable for this indication.

Background. With fast-growing evidence in literature for clinical applications of chemical exchange saturation transfer (CEST) magnetic resonance imaging (MRI), this prospective study aimed at applying amide proton transfer-weighted (APTw) CEST imaging in a clinical setting to assess its diagnostic potential in differentiation of intracranial tumors at 3 tesla (T). Methods. Using the asymmetry magnetization transfer ratio (MTRasym) analysis, CEST signals were quantitatively investigated in the tumor areas and in a similar sized region of the normal-appearing white matter (NAWM) on the contralateral hemisphere of 27 patients with intracranial tumors. Area under curve (AUC) analyses were used and results were compared to perfusion-weighted imaging (PWI). Results. Using APTw CEST, contrast-enhancing tumor areas showed significantly higher APTw CEST metrics than contralateral NAWM (AUC = 0.82; p < 0.01). In subgroup analyses of each tumor entity vs. NAWM, statistically significant effects were yielded for glioblastomas (AUC = 0.96; p < 0.01) and for meningiomas (AUC = 1.0; p < 0.01) but not for lymphomas as well as metastases (p > 0.05). PWI showed results comparable to APTw CEST in glioblastoma (p < 0.01). Conclusions. This prospective study confirmed the high diagnostic potential of APTw CEST imaging in a routine clinical setting to differentiate brain tumors.

BACKGROUND AND PURPOSE: After endovascular treatment of intracranial aneurysms with the Woven EndoBridge (WEB) device, worsening of aneurysm occlusion or re-opening was reported to be possibly associated with WEB shape modification. This study analyzed quantitatively the WEB shape modification with time in association with anatomic results. MATERIALS AND METHODS: Thirty patients with 32 WEB-treated intracranial aneurysms fulfilled the inclusion criteria of cranial CT at baseline (day of intervention) and a follow-up CT at least >1 months’ later (median follow-up time, 11.4 months; interquartile range, 6.5–21.5 months). Adequate occlusion was observed in 84.4%, and aneurysm remnant, in 15.6%. WEB shape modification was quantified by a semiautomated approach on CT scans. Time courses were evaluated graphically and analytically; association analyses were performed by linear mixed-effects regression models. RESULTS: In 29/32 WEB devices (90.6%), the reduction in height was found to be at least 5%. The decrease in height with time was significantly associated with the time interval in days since the intervention (P < .0001): A stronger decrease in WEB height was associated with a longer time since the intervention (median reduction in 1 year, 19.2%; range, 8.6%–52.3%). No significant association was found with the quality of aneurysm occlusion, device size, rupture status of the aneurysm, aneurysm location, and reintervention rate. CONCLUSIONS: Shape modification was common in WEB-treated intracranial aneurysms with a median height reduction of 19.2% in 1 year. The quality of aneurysm occlusion was not associated with WEB modification.

The timing of MRI imaging after surgical resection may have an important role in assessing the extent of resection (EoR) and in determining further treatment. The aim of our study was to evaluate the time dependency of T2 and FLAIR changes after surgery for LGG. The Log-Glio database of patients treated at our hospital from 2016 to 2021 was searched for patients >18a and non-enhancing intra-axial lesion with complete MR-imaging protocol. A total of 16 patients matched the inclusion criteria and were thus selected for volumetric analysis. All patients received an intraoperative scan (iMRI) after complete tumor removal, an ultra-early postoperative scan after skin closure, an early MRI within 48 h and a late follow up MRI after 3–4 mo. Detailed volumetric analysis of FLAIR and T2 abnormalities was conducted. Demographic data and basic characteristics were also analyzed. An ultra-early postoperative MRI was performed within a median time of 30 min after skin closure and showed significantly lower FLAIR (p = 0.003) and T2 (p = 0.003) abnormalities when compared to early postoperative MRI (median 23.5 h), though no significant difference was found between ultra-early and late postoperative FLAIR (p = 0.422) and T2 (p = 0.575) images. A significant difference was calculated between early and late postoperative FLAIR (p = 0.005) and T2 (p = 0.019) MRI scans. Additionally, we found no significant difference between intraoperative and ultra-early FLAIR/T2 (p = 0.919 and 0.499), but we found a significant difference between iMRI and early MRI FLAIR/T2 (p = 0.027 and p = 0.035). Therefore, a postoperative MRI performed 24 h or 48 h might lead to false positive findings. An MRI scan in the first hour after surgery (ultra-early) correlated best with residual tumor at 3 months follow up. An iMRI with open skull, at the end of resection, was similar to an ultra-early MRI with regard to residual tumor.