Uncertainty-aware deep learning for trustworthy prediction of long-term outcome after endovascular thrombectomy

Acute ischemic stroke (AIS) is a leading global cause of mortality and morbidity. Improving long-term outcome predictions after thrombectomy can enhance treatment quality by supporting clinical decision-making. With the advent of interpretable deep learning methods in recent years, it is now possible to develop trustworthy, high-performing prediction models. This study introduces an uncertainty-aware, graph deep learning model that predicts endovascular thrombectomy outcomes using clinical features and imaging biomarkers. The model targets long-term functional outcomes, defined by the three-month modified Rankin Score (mRS), and mortality rates. A sample of 220 AIS patients in the anterior circulation who underwent endovascular thrombectomy (EVT) was included, with 81 (37%) demonstrating good outcomes (mRS ≤ 2). The performance of the different algorithms evaluated was comparable, with the maximum validation under the curve (AUC) reaching 0.87 using graph convolutional networks (GCN) for mRS prediction and 0.86 using fully connected networks (FCN) for mortality prediction. Moderate performance was obtained at admission (AUC of 0.76 using GCN), which improved to 0.84 post-thrombectomy and to 0.89 a day after stroke. Reliable uncertainty prediction of the model could be demonstrated.

Accuracy of Dose-Saving Artificial-Intelligence-Based 3D Angiography (3DA) for Grading of Intracranial Artery Stenoses: Preliminary Findings

BACKGROUND AND PURPOSE

Based on artificial intelligence (AI), 3D angiography (3DA) is a novel postprocessing algorithm for "DSA-like" 3D imaging of cerebral vasculature. Because 3DA requires neither mask runs nor digital subtraction as the current standard 3D-DSA does, it has the potential to cut the patient dose by 50%. The object was to evaluate 3DA's diagnostic value for visualization of intracranial artery stenoses (IAS) compared to 3D-DSA.

MATERIALS AND METHODS

3D-DSA datasets of IAS (n_IAS = 10) were postprocessed using conventional and prototype software (Siemens Healthineers AG, Erlangen, Germany). Matching reconstructions were assessed by two experienced neuroradiologists in consensus reading, considering image quality (IQ), vessel diameters (VD_1/2), vessel-geometry index (VGI = VD₁/VD₂), and specific qualitative/quantitative parameters of IAS (e.g., location, visual IAS grading [low-/medium-/high-grade] and intra-/poststenotic diameters [d_{intra-/poststenotic} in mm]). Using the NASCET criteria, the percentual degree of luminal restriction was calculated.

RESULTS

In total, 20 angiographic 3D volumes (n_3DA = 10; n_3D-DSA = 10) were successfully reconstructed with equivalent IQ. Assessment of the vessel geometry in 3DA datasets did not differ significantly from 3D-DSA (VD₁: r = 0.994, p = 0.0001; VD₂:r = 0.994, p = 0.0001; VGI: r = 0.899, p = 0.0001). Qualitative analysis of IAS location (3DA/3D-DSA:n_ICA/C4 = 1, n_ICA/C7 = 1, n_MCA/M1 = 4, n_VA/V4 = 2, n_BA = 2) and the visual IAS grading (3DA/3D-DSA:n_low-grade = 3, n_medium-grade = 5, n_high-grade = 2) revealed identical results for 3DA and 3D-DSA, respectively. Quantitative IAS assessment showed a strong correlation regarding intra-/poststenotic diameters (r_{dintrastenotic} = 0.995, p_{dintrastenotic} = 0.0001; r_{dpoststenotic} = 0.995, p_{dpoststenotic} = 0.0001) and the percentual degree of luminal restriction (r_{NASCET 3DA} = 0.981; p_{NASCET 3DA} = 0.0001).

CONCLUSIONS

The AI-based 3DA is a resilient algorithm for the visualization of IAS and shows comparable results to 3D-DSA. Hence, 3DA is a promising new method that allows a considerable patient-dose reduction, and its clinical implementation would be highly desirable.

A differential of the left eye and right eye neurological pupil index is associated with discharge modified Rankin scores in neurologically injured patients

Background

Automated infrared pupillometry (AIP) and the Neurological Pupil index (NPi) provide an objective means of assessing and trending the pupillary light reflex (PLR) across a broad spectrum of neurological diseases. NPi quantifies the PLR and ranges from 0 to 5; in healthy individuals, the NPi of both eyes is expected to be ≥ 3.0 and symmetric. AIP values demonstrate emerging value as a prognostic tool with predictive properties that could allow practitioners to anticipate neurological deterioration and recovery. The presence of an NPi differential (a difference ≥ 0.7 between the left and right eye) is a potential sign of neurological abnormality.

Methods

We explored NPi differential by considering the modified Rankin Score at discharge (DC mRS) among patients admitted to neuroscience intensive care units (NSICU) of 4 U.S. and 1 Japanese hospitals and for two cohorts of brain injuries: stroke (including subarachnoid hemorrhage, intracerebral hemorrhage, acute ischemic stroke, and aneurysm, 1,200 total patients) and 185 traumatic brain injury (TBI) patients for a total of more than 54,000 pupillary measurements.

Results

Stroke patients with at least 1 occurrence of an NPi differential during their NSICU stay have higher DC mRS scores (3.9) compared to those without an NPi differential (2.7; P < .001). Patients with TBI and at least 1 occurrence of an NPi differential during their NSICU stay have higher discharge modified Rankin Scale scores (4.1) compared to those without an NPi differential (2.9; P < .001). When patients experience both abnormalities, abnormal (NPi < 3.0) and an NPi differential, the latter has an anticipatory relationship with respect to the former (P < .001 for z-score skewness analysis). Finally, our analysis confirmed ≥ 0.7 as the optimal cutoff value for the NPi differential (AUC = 0.71, P < .001).

Conclusion

The NPi differential is an important factor that clinicians should consider when managing critically ill neurological injured patients admitted to the neurocritical care units.

Trial registration

NCT02804438, Date of Registration: June 17, 2016.

Management von schwerwiegenden intrazerebralen Blutungen unter Antikoagulation

Die Inzidenz der intrazerebralen Blutung unter oraler Antikoagulation (OAK-ICB) steigt aufgrund der demografischen Entwicklung und der daraus resultierenden Zunahme an Patienten, die bei einem Vorhofflimmern antikoaguliert werden. Die OAK-ICB stellt eine der schwerwiegendsten Formen des hämorrhagischen Schlaganfalls dar, ist charakterisiert durch größere Blutungsvolumina, vermehrten Ventrikeleinbruch sowie häufigere und prolongierte Nachblutungen. Der Einfluss dieser Prognoseparameter mündet schließlich in eine erhöhte Sterblichkeit und ein hohes Maß an bleibender funktioneller Beeinträchtigung. Doch gerade bei diesem schweren Erkrankungsbild fehlt meist eine belastbare Evidenz, sodass die aktuellen internationalen und europäischen Leitlinien häufig nur schwache Therapieempfehlungen abgeben können. Im Rahmen der Akutbehandlung einer OAK-ICB ist das Hauptziel eine Minimierung der Hämatomprogression, welche durch Blutdruck- und Gerinnungsmanagement möglicherweise erreicht werden kann. Für alle OAK-ICBs erscheint die sofortige und aggressive Blutdrucksenkung auf einen systolischen Zielwert von 140 mmHg (mind. innerhalb von 4 – 6 Stunden) nach Aufnahme sinnvoll und für intrazerebrale Blutungen unter Vitamin-K-Antagonisten konnte eine Assoziation mit einer reduzierten Nachblutungsrate aufgezeigt werden. Die Antagonisierung der OAK-ICB sollte immer so rasch wie möglich, aber in Abhängigkeit von der jeweiligen oralen Antikoagulation erfolgen. Für die Vitamin-K-Antagonisten assoziierte ICB konnten große multizentrische Arbeiten in den letzten Jahren beitragen ein präziseres Vorgehen zu definieren. In Rahmen einer deutschlandweiten Kohortenstudie an knapp 1200 OAK-ICB (unter Marcumar bzw. Falithrom) konnten erstmals konkrete Zielwerte identifiziert werden. Eine Antagonisierung sollte mindestens bis zu einem INR-Wert < 1,3 bzw. 1,2 mindestens innerhalb von 4 Stunden nach Aufnahme erfolgen. Mittel der Wahl sollten Prothrombinkomplexkonzentrate (PPSB) sein, um eine rasche und vollständige Antagonisierung zu erzielen. Durch dieses kombinierte Procedere konnte eine deutliche Minimierung der Hämatomwachstumsraten und eine reduzierte Krankenhaussterblichkeit verzeichnet werden. Seit kurzem steht für Dabigatran assoziierte Blutungen ein Antidot (Idarucizumab) zur Verfügung, welches in Studien eine rasche und nahezu vollständige Elimination sowie suffiziente Hämostase aufzeigen konnte. Für Blutungen unter Faktor Xa-Inhibitoren besteht aktuell großer Handlungsbedarf, da bislang kein Antidot verfügbar ist und es ebenso unklar ist, ob in der klinischen Praxis eine relevante Antagonisierung durch gängige Substanzen, z. B. PPSB oder gefrorenes Frischplasma (GFP) erreicht werden kann. In dieser Übersichtsarbeit liegt der Schwerpunkt auf der Akut-Behandlung einer OAK-ICB und versucht anhand der aktuellen Datenlage konkrete Behandlungsoptionen aufzuzeigen. Ferner werden gegenwärtige Studien zur intensivmedizinischen Behandlung der ICB beleuchtet, um daraus relevante Implikationen für die Therapie einer OAK-ICB abzuleiten.

Acute Management of Anticoagulation-Associated Intracerebral Hemorrhage

The incidence of anticoagulation-associated intracerebral hemorrhage (OAC-ICH) is increasing along with demographic changes and patients requiring oral anticoagulation for atrial fibrillation. OAC-ICH is one of the most detrimental sub-types of hemorrhagic stroke, characterized by larger hemorrhage volumes, more frequent intraventricular hemorrhage, increased and prolonged hematoma enlargement. The impact of these outcome predictors’ results in greater mortality rates and larger numbers of patients left in functionally dependent states after ICH. Yet, evidence on treatment options for these severely affected patients remains limited, resulting in weak recommendations by international and European guidelines. Minimization of hematoma enlargement constitutes the main focus of acute care, possibly achieved by blood pressure and reversal management. In all OAC-ICH patients irrespective of anticoagulation agent, immediate and aggressive blood pressure reductions as fast as possible targeting systolic blood pressure levels of 140 mmHg (at least within 4–6 h) seems reasonable. Specifically, in vitamin-K-associated ICH, blood pressure reductions have been shown to be associated with reduced rates of hematoma enlargement. Reversal management in OAC-ICH should be carried out as quickly as possible, though accounting for the different anticoagulation agents used. In vitamin-K-associated ICH, recent large-sized multicenter investigations contributed to refining reversal strategies. One German-wide cohort study including roughly 1 200 patients identified for the first time beneficial target values to guide reversal treatment. Anticoagulation reversal should achieve INR values of less than 1.3 or 1.2 at least within 4 h after admission. Agents to be used comprise prothrombin complex concentrates (PCC), which may achieve faster and more complete reversal. This combined approach (blood pressure management and INR-reversal) was associated with significant reductions of hematoma enlargement and lower in-hospital mortality rates. For dabigatran-related bleeding complications, most recently an antidote (Idarucizumab) has been approved, which has been prospectively investigated and showed rapid and almost complete reversal as well as sufficient hemostasis. Currently, bleeding under the influence of factor-Xa inhibitors poses great difficulties because no antidote is available and efficacy of commonly used antagonizing agents, i. e., PCC or fresh-frozen plasma (FFP), has not been proven for clinical use. The main aim of the present review focuses on acute-care management of OAC-ICH and tries to depict treatment options based on the most current data. Moreover, current neurocritical-care therapies for ICH will be reviewed to identify relevant implications specific to OAC-ICH treatment.

No sex differences in long-term functional outcome after intracerebral hemorrhage

Background There is conflicting evidence about the influence of sex on outcome after spontaneous intracerebral hemorrhage (sICH) and the majority of the research focused on mortality and short-term outcome only. We investigated sex differences in long-term functional outcome after sICH. Methods We used data from a prospective hospital registry and included all consecutive patients with ICH admitted to our institution between January 2006 and July 2014. Functional outcome was assessed by modified Rankin Scale evaluated 3 and 12 months after ICH. We explored the influence of sex on long-term functional outcome using multivariable regression models and additionally by means of propensity score matching. Results We analyzed 823 patients, of whom 380 (46%) women. Women were on average three years older (p < 0.001), men had more often deep hematomas (p = 0.01). Unadjusted mortality rates were significantly increased in women at three months (42% vs. 35%; odds ratio (OR): 1.35; 95% confidence interval (CI): 1.02-1.80). After adjusting for baseline prognostic factors there were no differences between men and women in short- and long-term mortality (OR = 1.01; 95% CI = 0.66-1.54 and OR = 1.04; 95%CI = 0.69-1.57, respectively) and short- and long-term unfavorable outcome (OR = 1.02; 95%CI = 0.67-1.55 and OR = 0.96; 95% CI = 0.62-1.48, respectively). Conclusion We found no sex-related differences in long-term functional outcome in patients with sICH. The apparently worse functional outcome in women can be explained by differences in age.

Fibrinolysis Treatment for Cerebral Intraventricular Hemorrhage: A Temporal and Spatial Voxel-Based Analysis

OBJECTIVE

A voxel-based statistical approach on computer tomographic data in patients with intracerebral hemorrhage (ICH) and acute intraventricular hemorrhage (IVH) was used to evaluate spatial and temporal patterns of intraventricular blood in patients treated with intraventricular fibrinolysis (IVF) or without.

METHODS

IVH shapes were systematically assessed three dimensionally in patients with supratentorial ICH at three intervals of time (day of admission, day 4 ± 1, day 7+). The boundaries of the intraventricular blood clot were delineated on computed tomography (CT) scans using dedicated software. The CT scan and the IVH shape were transferred into stereotaxic space. In a second step, voxel-based statistics on group level were used to correlate the distribution of intraventricular blood with the interval and the treatment group.

RESULTS

Altogether 45 patients, 29 with IVF therapy and 16 without, were eligible to be included into this study. We found significant (false discovery rate [FDR] correction, q < .05) reduction of the intraventricular blood between day of admission and day 7 + for the third and fourth ventricle and parts of both lateral ventricles. In addition, we were able to show a significant difference between the IVF therapy and the conventionally treated group at day 4 ± 1 for the third ventricle.

CONCLUSIONS

The data indicate that voxel-based analysis on group level can be used to compare the time course and the distribution of intraventricular hemorrhage. This technique could be an interesting tool for future research on ICH with IVH.