Dataset Information

Outcome prediction in aneurysmal subarachnoid hemorrhage: a comparison of machine learning methods and established clinico-radiological scores.

ABSTRACT: Reliable prediction of outcomes of aneurysmal subarachnoid hemorrhage (aSAH) based on factors available at patient admission may support responsible allocation of resources as well as treatment decisions. Radiographic and clinical scoring systems may help clinicians estimate disease severity, but their predictive value is limited, especially in devising treatment strategies. In this study, we aimed to examine whether a machine learning (ML) approach using variables available on admission may improve outcome prediction in aSAH compared to established scoring systems. Combined clinical and radiographic features as well as standard scores (Hunt & Hess, WFNS, BNI, Fisher, and VASOGRADE) available on patient admission were analyzed using a consecutive single-center database of patients that presented with aSAH (n = 388). Different ML models (seven algorithms including three types of traditional generalized linear models, as well as a tree bosting algorithm, a support vector machine classifier (SVMC), a Naive Bayes (NB) classifier, and a multilayer perceptron (MLP) artificial neural net) were trained for single features, scores, and combined features with a random split into training and test sets (4:1 ratio), ten-fold cross-validation, and 50 shuffles. For combined features, feature importance was calculated. There was no difference in performance between traditional and other ML applications using traditional clinico-radiographic features. Also, no relevant difference was identified between a combined set of clinico-radiological features available on admission (highest AUC 0.78, tree boosting) and the best performing clinical score GCS (highest AUC 0.76, tree boosting). GCS and age were the most important variables for the feature combination. In this cohort of patients with aSAH, the performance of functional outcome prediction by machine learning techniques was comparable to traditional methods and established clinical scores. Future work is necessary to examine input variables other than traditional clinico-radiographic features and to evaluate whether a higher performance for outcome prediction in aSAH can be achieved.

SUBMITTER: Dengler NF

PROVIDER: S-EPMC8490233 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Aneurysmal subarachnoid hemorrhage is a life-threatening disease requiring neurocritical care. Delayed cerebral ischemia is a well-known complication that contributes to unfavorable neurological outcomes. Cerebral vasospasm has been thought to be the main cause of delayed cerebral ischemia, and although several studies were able to decrease cerebral vasospasm, none showed improved neurological outcomes. Our target is not cerebral vasospasm but improving neurological outcomes. The purpose of this review is to discuss what intensivists should know and can do to improve clinical outcomes in subarachnoid hemorrhage patients.Delayed cerebral ischemia is thought to be due to not only vasospasm but also multifactorial mechanisms. Additionally, the concept of early brain injury, which occurs within the first 72 h after the hemorrhage, has become an important concern. Increasing sympathetic activity after the hemorrhage is associated with cardiopulmonary complications and poor outcomes. Serum lactate measurement may be a valuable marker reflecting the severity of sympathetic activity. The transpulmonary thermodilution method will bring about an advanced understanding of hemodynamic management. Fever is a well-recognized symptom and targeted temperature management is an anticipated intervention. To avoid hyperglycemia and hypoglycemia, performing moderate glucose control and minimizing glucose variability are important concepts in glycemic management, but the optimal target range remains unknown. Dysnatremia seems to be associated with negative outcomes. It is not clear yet that maintaining normonatremia actively improves neurological outcomes. Optimal duration of intensive care management has not been determined.Although we have an advanced understanding of the pathophysiology and clinical characteristics of subarachnoid hemorrhage, there are many controversies in the intensive care unit management of subarachnoid hemorrhage. With an awareness of not only delayed cerebral ischemia but also early brain injury, more attention should be given to various aspects to improve neurological outcomes.

Project description:BackgroundDisease progression and delayed neurological complications are common after aneurysmal subarachnoid hemorrhage (aSAH). We explored the potential of quantitative blood-brain barrier (BBB) imaging to predict disease progression and neurological outcome.MethodsData were collected as part of the Co-Operative Studies of Brain Injury Depolarizations (COSBID). We analyzed retrospectively, blinded and semi-automatically magnetic resonance images from 124 aSAH patients scanned at 4 time points (24-48 h, 6-8 days, 12-15 days and 6-12 months) after the initial hemorrhage. Volume of brain with apparent pathology and/or BBB dysfunction (BBBD), subarachnoid space and lateral ventricles were measured. Neurological status on admission was assessed using the World Federation of Neurosurgical Societies and Rosen-Macdonald scores. Outcome at ≥6 months was assessed using the extended Glasgow outcome scale and disease course (progressive or non-progressive based on imaging-detected loss of normal brain tissue in consecutive scans). Logistic regression was used to define biomarkers that best predict outcomes. Receiver operating characteristic analysis was performed to assess accuracy of outcome prediction models.FindingsIn the present cohort, 63% of patients had progressive and 37% non-progressive disease course. Progressive course was associated with worse outcome at ≥6 months (sensitivity of 98% and specificity of 97%). Brain volume with BBBD was significantly larger in patients with progressive course already 24-48 h after admission (2.23 (1.23-3.17) folds, median with 95%CI), and persisted at all time points. The highest probability of a BBB-disrupted voxel to become pathological was found at a distance of ≤1 cm from the brain with apparent pathology (0·284 (0·122-0·594), p < 0·001, median with 95%CI). A multivariate logistic regression model revealed power for BBBD in combination with RMS at 24-48 h in predicting outcome (ROC area under the curve = 0·829, p < 0·001).InterpretationWe suggest that early identification of BBBD may serve as a key predictive biomarker for neurological outcome in aSAH. FUND: Dr. Dreier was supported by grants from the Deutsche Forschungsgemeinschaft (DFG) (DFG DR 323/5-1 and DFG DR 323/10-1), the Bundesministerium für Bildung und Forschung (BMBF) Center for Stroke Research Berlin 01 EO 0801 and FP7 no 602150 CENTER-TBI. Dr. Friedman was supported by grants from Israel Science Foundation and Canada Institute for Health Research (CIHR). Dr. Friedman was supported by grants from European Union's Seventh Framework Program (FP7/2007-2013; grant #602102).

Dataset Information

Outcome prediction in aneurysmal subarachnoid hemorrhage: a comparison of machine learning methods and established clinico-radiological scores.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets