Labeling Noncontrast Head CT Reports for Common Findings Using Natural Language Processing.

BACKGROUND AND PURPOSE: Prioritizing reading of noncontrast head CT examinations through an automated triage system may improve time to care for patients with acute neuroradiologic findings. We present a natural language-processing approach for labeling findings in noncontrast head CT reports, which permits creation of a large, labeled dataset of head CT images for development of emergent-finding detection and reading-prioritization algorithms. MATERIALS AND METHODS: In this retrospective study, 1002 clinical radiology reports from noncontrast head CTs collected between 2008 and 2013 were manually labeled across 12 common neuroradiologic finding categories. Each report was then encoded using an n-gram model of unigrams, bigrams, and trigrams. A logistic regression model was then trained to label each report for every common finding. Models were trained and assessed using a combination of L2 regularization and 5-fold cross-validation. RESULTS: Model performance was strongest for the fracture, hemorrhage, herniation, mass effect, pneumocephalus, postoperative status, and volume loss models in which the area under the receiver operating characteristic curve exceeded 0.95. Performance was relatively weaker for the edema, hydrocephalus, infarct, tumor, and white-matter disease models (area under the receiver operating characteristic curve > 0.85). Analysis of coefficients revealed finding-specific words among the top coefficients in each model. Class output probabilities were found to be a useful indicator of predictive error on individual report examples in higher-performing models. CONCLUSIONS: Combining logistic regression with n-gram encoding is a robust approach to labeling common findings in noncontrast head CT reports.

Consensus summary statement of the International Multidisciplinary Consensus Conference on Multimodality Monitoring in Neurocritical Care: a statement for healthcare professionals from the Neurocritical Care Society and the European Society of Intensive Care Medicine

Neurocritical care depends, in part, on careful patient monitoring but as yet there are little data on what processes are the most important to monitor, how these should be monitored, and whether monitoring these processes is cost-effective and impacts outcome. At the same time, bioinformatics is a rapidly emerging field in critical care but as yet there is little agreement or standardization on what information is important and how it should be displayed and analyzed. The Neurocritical Care Society in collaboration with the European Society of Intensive Care Medicine, the Society for Critical Care Medicine, and the Latin America Brain Injury Consortium organized an international, multidisciplinary consensus conference to begin to address these needs. International experts from neurosurgery, neurocritical care, neurology, critical care, neuroanesthesiology, nursing, pharmacy, and informatics were recruited on the basis of their research, publication record, and expertise. They undertook a systematic literature review to develop recommendations about specific topics on physiologic processes important to the care of patients with disorders that require neurocritical care. This review does not make recommendations about treatment, imaging, and intraoperative monitoring. A multidisciplinary jury, selected for their expertise in clinical investigation and development of practice guidelines, guided this process. The GRADE system was used to develop recommendations based on literature review, discussion, integrating the literature with the participants' collective experience, and critical review by an impartial jury. Emphasis was placed on the principle that recommendations should be based on both data quality and on trade-offs and translation into clinical practice. Strong consideration was given to providing pragmatic guidance and recommendations for bedside neuromonitoring, even in the absence of high quality data

Predictors of hemorrhage volume and disability after perimesencephalic subarachnoid hemorrhage.

OBJECTIVE: The determinants of subarachnoid hemorrhage (SAH) volume and an atypical pattern of blood are not clear. Our objective was to determine if reduced platelet activity on admission and abnormal venous drainage are associated with greater SAH volume. METHODS: We prospectively identified noncomatose patients with SAH without an identifiable aneurysm. We routinely measured platelet activity on admission and recorded aspirin use. SAH volumes were calculated with a validated technique. CT angiograms were reviewed by a certified neuroradiologist for venous drainage. Patients were followed for clinical outcomes through 3 months with the modified Rankin Scale (mRS). Data are Q1-Q3. RESULTS: There were 31 patients in the cohort. Thirty (97%) underwent an angiogram on admission, and 25 (81%) an additional delayed angiogram. SAH volume was lowest with normal venous drainage bilaterally (4.4 [3.7-16.4] mL) and higher with 1 (12.9 [3.7-20.4]) or 2 (20.9 [12.5-34.6] mL, p = 0.03) discontinuous venous drainages. Patients with reduced platelet activity had more SAH on the diagnostic CT (17.5 [10.6-20.9] vs 6.1 [2.3-15.3] mL) (p = 0.046). SAH volume was greater for patients requiring drainage for hydrocephalus (16.4 [11.5-20.5] vs 5.4 [2.7-16.4] mL) (p = 0.009). Outcomes at 3 months were generally excellent (median mRS = 0, no symptoms). CONCLUSIONS: Discontinuous venous drainage and reduced platelet activity were associated with increased SAH volume and hydrocephalus. These factors may explain thick SAH and reduce the need for repeated invasive imaging in such patients.

Packed red blood cell age does not impact adverse events or outcomes after subarachnoid haemorrhage.

OBJECTIVES: To determine if the age of packed red blood cells (PRBCs) is associated with adverse events or outcomes in patients with subarachnoid haemorrhage (SAH). AIM: Analyse data on PRBC age to complications and outcomes in patients with SAH. BACKGROUND: Patients who receive a PRBC transfusion after SAH have a higher rate of complications, and older PRBC age may be responsible for this. METHODS/MATERIALS: We prospectively recorded clinical and demographic data, acute adverse effects related to transfusion, major hospital events, radiographic cerebral infarction, PRBC age and outcomes in 119 patients with SAH who received a PRBC transfusion. Patients were followed for outcomes at 14 days or discharge, 28 days and 3 months with the modified Rankin scale (a measure of neurologic function). RESULTS: In 241 PRBC transfusions, there was new fever in 36 (15%), hypotension in 23 (10%), pulmonary oedema or symptomatic respiratory distress in 5 (2%) and rash in 1 (1%). Age of PRBCs administered was not associated with vasospasm, cerebral infarction, acute adverse events or outcomes (P > 0·1 for all). CONCLUSIONS: In this small registry of patients with SAH, the age of transfused PRBCs was not associated with adverse events or outcomes.

Acute physiological derangement is associated with early radiographic cerebral infarction after subarachnoid haemorrhage.

BACKGROUND: Cerebral infarction after aneurysmal subarachnoid haemorrhage (SAH) is presumed to be due to cerebral vasospasm, defined as arterial lumen narrowing from days 3 to 14. METHODS: We reviewed the computed tomography scans of 103 patients with aneurysmal SAH for radiographic cerebral infarction and controlled for other predictors of outcome. A blinded neuroradiologist reviewed the angiograms. Cerebral infarction from vasospasm was judged to be unlikely if it was visible on computed tomography within 2 calendar days of SAH or if angiography showed no vasospasm in a referable vessel, or both. RESULTS: Cerebral infarction occurred in 29 (28%) of 103 patients with SAH. 18 patients had cerebral infarction that was unlikely to be due to vasospasm because it was visible on computed tomography by day 2 (6 (33%)) or because angiography showed no vasospasm in a referable artery (7 (39%)), or both (5 (28%)). In a multivariate model, cerebral infarction was significantly related to World Federation of Neurologic Surgeons grade (odds ratio (OR) 1.5/grade, 95% confidence interval (CI) 1.1 to 2.01, p = 0.006) and SAH-Physiologic Derangement Score (PDS) >2 (OR 3.7, 95% CI 1.4 to 9.8, p = 0.01) on admission. Global cerebral oedema (OR 4.3, 95% CI 1.5 to 12.5, p = 0.007) predicted cerebral infarction. Patients with cerebral infarction detectable by day 2 had a higher SAH-PDS than patients with later cerebral infarction (p = 0.025). CONCLUSIONS: Many cerebral infarctions after SAH are unlikely to be caused by vasospasm because they occur too soon after SAH or because angiography shows no vasospasm in a referable artery, or both. Physiological derangement and cerebral oedema may be worthwhile targets for intervention to decrease the occurrence and clinical impact of cerebral infarction after SAH.