Automatic Prediction of Infarct Growth in Acute Ischemic Stroke from MR Apparent Diffusion Coefficient Maps1

Rationale and Objectives

We introduce a new approach to the prediction of final infarct growth in human acute ischemic stroke based on image analysis of the apparent diffusion coefficient (ADC) maps obtained from magnetic resonance imaging. Evidence from multiple previous studies indicate that ADC maps are likely to reveal brain regions belonging to the ischemic penumbra, that is, areas that may be at risk of infarction in the few hours following stroke onset.

Materials and Methods

In a context where “time is brain,” and contrarily to the alternative—and still-debated—perfusion-diffusion weighted image (PWI/DWI) mismatch approach, the DWI magnetic resonance sequences are standardized, fast to acquire, and do not necessitate injection of a contrast agent. The image analysis approach presented here consists of the segmentation of the ischemic penumbra using a fast three-dimensional region-growing technique that mimics the growth of the infarct lesion during acute stroke.

Results

The method was evaluated with both numerical simulations and on two groups of 20 ischemic stroke patients (40 patients total). The first group of patient data was used to adjust the parameters of the model ruling the region-growing procedure. The second group of patient data was dedicated to evaluation purposes only, with no subsequent adjustment of the free parameters of the image-analysis procedure. Results indicate that the predicted final infarct volumes are significantly correlated with the true final lesion volumes as revealed by follow-up measurements from DWI sequences.

Conclusion

The DWI-ADC mismatch method is an encouraging fast alternative to the PWI-DWI mismatch approach to evaluate the likeliness of infarct growth during the acute stage of ischemic stroke.

Key Words: Stroke, magnetic resonance imaging, diffusion imaging, region-growing segmentation, cerebral ischemia, infarct growth prediction, apparent diffusion coefficient