Comparison of classification methods for tissue outcome after ischaemic stroke

• Published in: The European journal of neuroscience 2019-11, Vol.50 (10), p.3590-3598
• Main Authors: Tozlu, Ceren, Ozenne, Brice, Cho, Tae‐Hee, Nighoghossian, Norbert, Mikkelsen, Irene Klærke, Derex, Laurent, Hermier, Marc, Pedraza, Salvador, Fiehler, Jens, Østergaard, Leif, Berthezène, Yves, Baron, Jean‐Claude, Maucort‐Boulch, Delphine
• Format: Article
• Language: English
• Subjects: brain ischaemia; Cerebral infarction; Classification; Decision making; diffusion‐weighted imaging; Learning algorithms; Life Sciences; machine learning; Magnetic resonance imaging; Neural networks; Neuroimaging; NMR; Nuclear magnetic resonance; Perfusion; perfusion‐weighted imaging; Statistical analysis; Stroke
• ISSN: 0953-816X; 1460-9568
• DOI: 10.1111/ejn.14507

In acute ischaemic stroke, identifying brain tissue at high risk of infarction is important for clinical decision‐making. This tissue may be identified with suitable classification methods from magnetic resonance imaging data. The aim of the present study was to assess and compare the performance of five popular classification methods (adaptive boosting, logistic regression, artificial neural networks, random forest and support vector machine) in identifying tissue at high risk of infarction on human voxel‐based brain imaging data. The classification methods were used with eight MRI parameters, including diffusion‐weighted imaging and perfusion‐weighted imaging obtained in 55 patients. The five criteria used to assess the performance of the methods were the area under the receiver operating curve (AUCroc), the area under the precision–recall curve (AUCpr), sensitivity, specificity and the Dice coefficient. The methods performed equally in terms of sensitivity and specificity, while the results of AUCroc and the Dice coefficient were significantly better for adaptive boosting, logistic regression, artificial neural networks and random forest. However, there was no statistically significant difference between the performances of these five classification methods regarding AUCpr, which was the main comparison metric. Machine learning methods can provide valuable prognostic information using multimodal imaging data in acute ischaemic stroke, which in turn can assist in developing personalized treatment decision for clinicians after a thorough validation of methods with an independent data set. The machine learning methods used in our study (adaptive boosting, logistic regression, artificial neural networks, random forest and support vector machine) gave equivalently satisfactory results in distinguishing the healthy voxels from the infarcted voxels in stroke patients. The machine learning methods implemented in our study may provide a valuable tool that can assist for the treatment decision after a thorough validation of the methods using a larger, multi‐model and independent data set.