Machine Learning in Acute Ischemic Stroke Neuroimaging

Under current guidelines, intravenous thrombolysis is used to treat acute stroke only if it can be ascertained that the time since the onset of symptoms was less than 4.5 hours. We sought to determine whether patients with stroke with an unknown time of onset and features suggesting recent cerebral infarction on magnetic resonance imaging (MRI) would benefit from thrombolysis with the use of intravenous alteplase.

Randomized clinical trials suggest the benefit of intravenous tissue-type plasminogen activator (tPA) in acute ischemic stroke is time dependent. However, modest sample sizes have limited characterization of the extent to which onset to treatment (OTT) time influences outcome; and the generalizability of findings to clinical practice is uncertain. To evaluate the degree to which OTT time is associated with outcome among patients with acute ischemic stroke treated with intraveneous tPA. Data were analyzed from 58,353 patients with acute ischemic stroke treated with tPA within 4.5 hours of symptom onset in 1395 hospitals participating in the Get With The Guidelines-Stroke Program, April 2003 to March 2012. Relationship between OTT time and in-hospital mortality, symptomatic intracranial hemorrhage, ambulatory status at discharge, and discharge destination. Among the 58,353 tPA-treated patients, median age was 72 years, 50.3% were women, median OTT time was 144 minutes (interquartile range, 115-170), 9.3% (5404) had OTT time of 0 to 90 minutes, 77.2% (45,029) had OTT time of 91 to 180 minutes, and 13.6% (7920) had OTT time of 181 to 270 minutes. Median pretreatment National Institutes of Health Stroke Scale documented in 87.7% of patients was 11 (interquartile range, 6-17). Patient factors most strongly associated with shorter OTT included greater stroke severity (odds ratio [OR], 2.8; 95% CI, 2.5-3.1 per 5-point increase), arrival by ambulance (OR, 5.9; 95% CI, 4.5-7.3), and arrival during regular hours (OR, 4.6; 95% CI, 3.8-5.4). Overall, there were 5142 (8.8%) in-hospital deaths, 2873 (4.9%) patients had intracranial hemorrhage, 19,491 (33.4%) patients achieved independent ambulation at hospital discharge, and 22,541 (38.6%) patients were discharged to home. Faster OTT, in 15-minute increments, was associated with reduced in-hospital mortality (OR, 0.96; 95% CI, 0.95-0.98; P < .001), reduced symptomatic intracranial hemorrhage (OR, 0.96; 95% CI, 0.95-0.98; P < .001), increased achievement of independent ambulation at discharge (OR, 1.04; 95% CI, 1.03-1.05; P < .001), and increased discharge to home (OR, 1.03; 95% CI, 1.02-1.04; P < .001). In a registry representing US clinical practice, earlier thrombolytic treatment was associated with reduced mortality and symptomatic intracranial hemorrhage, and higher rates of independent ambulation at discharge and discharge to home following acute ischemic stroke. These findings support intensive efforts to accelerate hospital presentation and thrombolytic treatment in patients with stroke.

The objective of this study is to investigate the use of pattern classification methods for distinguishing different types of brain tumors, such as primary gliomas from metastases, and also for grading of gliomas. The availability of an automated computer analysis tool that is more objective than human readers can potentially lead to more reliable and reproducible brain tumor diagnostic procedures. A computer-assisted classification method combining conventional MRI and perfusion MRI is developed and used for differential diagnosis. The proposed scheme consists of several steps including region-of-interest definition, feature extraction, feature selection, and classification. The extracted features include tumor shape and intensity characteristics, as well as rotation invariant texture features. Feature subset selection is performed using support vector machines with recursive feature elimination. The method was applied on a population of 102 brain tumors histologically diagnosed as metastasis (24), meningiomas (4), gliomas World Health Organization grade II (22), gliomas World Health Organization grade III (18), and glioblastomas (34). The binary support vector machine classification accuracy, sensitivity, and specificity, assessed by leave-one-out cross-validation, were, respectively, 85%, 87%, and 79% for discrimination of metastases from gliomas and 88%, 85%, and 96% for discrimination of high-grade (grades III and IV) from low-grade (grade II) neoplasms. Multiclass classification was also performed via a one-vs-all voting scheme. (c) 2009 Wiley-Liss, Inc.