Expert-augmented automated machine learning optimizes hemodynamic predictors of spinal cord injury outcome.

Artificial intelligence and machine learning (AI/ML) is becoming increasingly more accessible to biomedical researchers with significant potential to transform biomedicine through optimization of highly-accurate predictive models and enabling better understanding of disease biology. Automated machine learning (AutoML) in particular is positioned to democratize artificial intelligence (AI) by reducing the amount of human input and ML expertise needed. However, successful translation of AI/ML in biomedicine requires moving beyond optimizing only for prediction accuracy and towards establishing reproducible clinical and biological inferences. This is especially challenging for clinical studies on rare disorders where the smaller patient cohorts and corresponding sample size is an obstacle for reproducible modeling results. Here, we present a model-agnostic framework to reinforce AutoML using strategies and tools of explainable and reproducible AI, including novel metrics to assess model reproducibility. The framework enables clinicians to interpret AutoML-generated models for clinical and biological verifiability and consequently integrate domain expertise during model development. We applied the framework towards spinal cord injury prognostication to optimize the intraoperative hemodynamic range during injury-related surgery and additionally identified a strong detrimental relationship between intraoperative hypertension and patient outcome. Furthermore, our analysis captured how evolving clinical practices such as faster time-to-surgery and blood pressure management affect clinical model development. Altogether, we illustrate how expert-augmented AutoML improves inferential reproducibility for biomedical discovery and can ultimately build trust in AI processes towards effective clinical integration.

Smaller left anterior cingulate cortex in non-bipolar relatives of patients with bipolar disorder

Objective: Bipolar disorder (BD) is highly heritable. The present study aimed at identifying brain morphometric features that could represent markers of BD vulnerability in non-bipolar relatives of bipolar patients. Methods: In the present study, structural magnetic resonance imaging brain scans were acquired from a total of 93 subjects, including 31 patients with BD, 31 non-bipolar relatives of BD patients, and 31 healthy controls. Volumetric measurements of the anterior cingulate cortex (ACC), lateral ventricles, amygdala, and hippocampus were completed using the automated software FreeSurfer. Results: Analysis of covariance (with age, gender, and intracranial volume as covariates) indicated smaller left ACC volumes in unaffected relatives as compared to healthy controls and BD patients (p = 0.004 and p = 0.037, respectively). No additional statistically significant differences were detected for other brain structures. Conclusion: Our findings suggest smaller left ACC volume as a viable biomarker candidate for BD.

Rapidly-progressive catatonia responsive to zolpidem in a patient with ovarian teratoma-associated paraneoplastic encephalitis.

Psychiatric symptoms and catatonia are key components of the clinical presentation of paraneoplastic encephalitis; additionally symptoms can be long-lasting and often difficult to treat. We report a 73-year-old patient with rapidly progressive catatonia not responsive to immunotherapy, tumor resection, electroconvulsive therapy, or benzodiazepines who had significant improvement after zolpidem administration. This report suggests that zolpidem is an option in the treatment of patients with refractory catatonia and paraneoplastic encephalitis.

Malignant EEG patterns in cardiac arrest patients treated with targeted temperature management who survive to hospital discharge.

BACKGROUND AND PURPOSE: Cardiac arrest patients treated with targeted temperature management (TTM) have improved neurological outcomes, however mortality remains high. EEG monitoring improves detection of malignant EEG patterns (MEPs), however their prevalence in patients surviving to hospital discharge is unknown. DESIGN/METHODS: We examined consecutive cardiac arrest subjects who received TTM and continuous EEG monitoring at one academic center. Only subjects surviving to hospital discharge were included in the analysis. MEPs were defined as seizures, status epilepticus, myoclonic status epilepticus, or generalized periodic discharges. Subjects with suppression-burst (SB) without concomitant MEPs were categorized as having a "pure" SB pattern. Demographic, survival, hospital discharge disposition, and neurological function data were recorded retrospectively. Outcomes were assessed using the Glasgow-Pittsburgh Cerebral Performance Category (CPC). A CPC score of 1-2 was considered "good" neurological function, and a CPC of 3-4 "poor". RESULTS: Of 364 admissions due to cardiac arrest screened, 120 (29.9%) survived to hospital discharge and met inclusion criteria. MEPs and pure SB were observed in 19 (15.8%) and 22 (18.3%) survivors respectively. Two subjects with MEP and eight subjects with pure SB had good neurological function at discharge, however all SB cases were confounded by the use of anesthetic agents. Presence of MEPs was not an independent predictor of poor neurological function (p=0.1). CONCLUSIONS: MEPs are common among cardiac arrest patients treated with induced hypothermia who survive to hospital discharge. Poor neurological function at discharge was not associated with MEPs. Prospective studies assessing the role of EEG monitoring in cardiac arrest prognostication are warranted.

Outcomes after endovascular treatment for anterior circulation stroke presenting as wake-up strokes are not different than those with witnessed onset beyond 8†hours.

OBJECTIVE: Previous studies have suggested that patients with wake-up stroke (WUS) may have superior outcomes compared with patients with a witnessed late time of onset after revascularization. We sought to test this hypothesis in patients with anterior circulation large vessel occlusion stroke (ACLVOS) treated with endovascular therapy beyond 8 h from time last seen well (TLSW). METHODS: A single center retrospective review of a prospectively acquired database of consecutive patients was performed to identify patients presenting beyond 8 h of TLSW with radiographic evidence of ACLVOS, small core, and large penumbra who subsequently underwent endovascular treatment. RESULTS: We identified 206 patients. Patients were divided into two groups: (1) patients with WUS (38%, n=78) and (2) patients with witnessed onset beyond 8 h (62%, n=128). The groups were similar in age, baseline National Institutes of Health Stroke Scale score, TLSW to reperfusion, baseline infarct volume, and rate of successful recanalization. Rates of good outcome (modified Rankin Scale score of 0-2 at 90 days, 43% vs. 50%, p=0.3), parenchymal hematoma (9% vs. 5.5%, p=0.3), and final infarct volume (75.2 vs. 61.4 mL, p=0.6) were comparable. Multivariate analysis identified age (OR=0.95, 95% CI 0.91 to 0.99, p<0.042), successful recanalization (OR 6.0, 95% CI 1.5 to 23.5, p=0.009), and final infarct volume (OR 0.98, 95% CI 0.97 to 0.99, p<0.001) but not mode of presentation as predictors of favorable outcomes. CONCLUSIONS: Rates of good outcomes, parenchymal hematoma, and final infarct volumes following endovascular treatment may not be different in patients with WUS compared with patients with witnessed onset of symptoms beyond 8 h.