An automated quality control (QC) system is essential to ensure streamlined head computed tomography (CT) scan interpretations that do not affect subsequent image analysis. Such a system is advantageous compared to current human QC protocols, which are subjective and time-consuming. In this work, we aim to develop a deep learning-based framework to classify a scan to be of usable or unusable quality. Supervised deep learning models have been highly effective in classification tasks, but they are highly complex and require large, annotated data for effective training. Additional challenges with QC datasets include -1) class-imbalance -usable cases far exceed the unusable ones and 2) weak-labels -scan level labels may not match slice level labels. The proposed framework utilizes these weak labels to augment a standard anomaly detection technique. Specifically, we proposed a hybrid model that consists of a variational autoencoder (VAE) and a Siamese Neural Network (SNN). While the VAE is trained to learn how usable scans appear and reconstruct an input scan, the SNN compares how similar this input scan is to its reconstruction and flags the ones that are less similar than a threshold. The proposed method is more suited to capture the differences in non-linear feature structure between the two classes of data than typical anomaly detection methods that depend on intensity-based metrics like root mean square error (RMSE). Comparison with state-of-the-art anomaly detection methods using multiple classification metrics establishes superiority of the proposed framework in flagging inferior quality scans for review by radiologists, thus reducing their workload and establishing a reliable and consistent dataflow.

Background: Optimal level blood pressure (BP) targets in acute stroke remain elusive. Tailored hemodynamic management after endovascular thrombectomy (EVT) may reduce the risk of reperfusion injury and promote penumbral recovery. Our study aimed to evaluate the relationship between personalized autoregulation-based BP targets, secondary brain injury, and functional outcomes. Methods: We prospectively enrolled 200 patients with acute ischemic stroke who underwent EVT. Autoregulatory function was continuously measured for >=24 hours using simultaneous recordings of near-infrared spectroscopy and mean arterial pressure (MAP). The resulting autoregulatory index was used to calculate and trend the BP range at which autoregulation was most preserved. Percent time and “dose” that MAP exceeded the upper limit or dropped below the lower limit of autoregulation (ULA, LLA) were calculated for each patient. Hemodynamic parameters were correlated with short-term clinical endpoints (symptomatic ICH), biomarkers of secondary brain injury (net water uptake, hemorrhagic transformation (HT), infarct progression), and 90-day functional outcomes. Results: Personalized BP targets were successfully computed in 195 patients (mean age 70 ± 16, 45% female, mean NIHSS 14, mean monitoring time 31 ± 28 hours). Time above the ULA was associated with worse functional outcomes at 90-days after adjusting for age, sex, NIHSS, ASPECTS and TICI (adjusted OR per 10% increase 1.4, 95% CI 1.1-1.6, P=0.004). The burden of hyperperfusion was significantly greater among patients with HT (median 2.7 vs. 3.2 mmHg*min, p=0.01) and sICH (median 2.8 vs. 4.8 mmHg*min, p=0.05) than in those without it. Furthermore, time spent above the ULA was significantly correlated with net water uptake at 72 hours (r=0.37, p=0.03). Among patients with unsuccessful reperfusion, there was a non-significant correlation between time below the LLA and infarct progression (r=0.35, p=0.064). Conclusions: In the largest study conducted to date, deviations from personalized BP targets were associated with an increased risk of secondary brain injury and worse functional outcomes. Autoregulation-oriented BP management represents a promising strategy for maximizing recovery after ischemic stroke.

Introduction: High blood pressure after endovascular thrombectomy (EVT) can cause cerebral hyperemia and disrupt the blood-brain barrier. However, its role in cerebral edema development is incompletely understood. In this study, we examined the relationship between post-EVT systolic blood pressure (SBP) trajectories and cerebral edema. Methods: We prospectively enrolled patients with large-vessel occlusion stroke who underwent EVT. Cerebrospinal fluid (CSF) volume was measured using a deep-learning algorithm on CT images at baseline, 24 hours, and 72 hours after stroke. The ratio of CSF volumes between hemispheres was calculated. Automated segmentation of infarct regions on follow-up scans was used to measure net water uptake (NWU), the ratio of density within infarcted tissue relative to the mirrored contralateral region. Latent variable mixture modeling (LVMM) divided patients into SBP trajectory groups during the first 72 hours post-EVT (Fig. 1A). Measures of edema (change in CSF ratio, NWU) were compared between groups. Results: One hundred patients (mean age 70 ± 16, mean NIHSS 15) were analyzed. Edema was assessed by a gradual increase in NWU (20.5, 27.0) at 24 and 72 hours, respectively, and by a reduction in CSF ratio (0.95, 0.78, 0.68) in the affected hemisphere at baseline, 24 hours, and 72 hours, respectively. LVMM identified five SBP trajectories. Higher SBP trajectories were associated with higher NWU (Fig. 1B) but not lower CSF ratio at 24 hours (p<0.001 and p=0.343, respectively). After adjusting for age, admission NIHSS, and TICI score, the moderate-to-high and high-to-moderate trajectory groups were independently associated with higher NWU (aOR 11.40, 95% CI 2.14-20.66) and (aOR 10.97, 95% CI 0.12-21.82), relative to the low and moderate groups. Conclusions: Higher SBP trajectories are associated with an increase in NWU post-EVT. NWU is a promising radiographic biomarker for measuring cerebral edema during the early phase after stroke.

No abstract