Radiology: Time is brainMRI may predict stroke onset, extend treatment
 | |
| Image source: Cedars-Sinai Medical Center in Los Angeles |
Currently, intravenous thrombolysis is recommended to break down clots in ischemic stroke patients. However, many patients who would benefit from thrombolysis treatment are ineligible for the therapy because their time of stroke onset is unknown.
"This is a major concern for a large number of patients who are currently a priori excluded from thrombolysis treatment. For instance, in up to one quarter of patients, acute stroke symptoms are detected upon awakening, while in other patients the symptoms are unwitnessed or there may be general uncertainty concerning the precise time of stroke onset," explained the study's senior author Catherine Oppenheim, MD, PhD, of the department of radiology and neurology at the University of Paris Descartes.
The authors performed a retrospective study of 130 patients admitted for acute ischemic stroke with known stroke onset times and National Institute of Health Stroke Scale scores of greater than three. All patients were imaged with 1.5T MRI within 12 hours of stroke using fluid-attenuated inversion recovery (FLAIR), diffusion-weighted (DW) imaging and apparent diffusion coefficient (ADC) measurements. Two independent readers estimated stroke onset based on these three parameters.
Sixty-three of the 130 patients involved in the study were imaged within three hours of stroke onset (hyperacute group), while the remaining 67 patients were imaged between three and 12 hours after the start of stroke (acute group). All patients treated with thrombolysis were administered the therapy after MRI (so images were unaffected by treatment).
Of the 130 patients imaged, 123 were correctly categorized into either the hyperacute or acute groups. Interobserver agreement on classifying patients "was almost perfect," with a kappa value of .97. Four patients imaged within three hours of stroke were wrongly assigned to the acute group, while three acute patients were wrongly categorized as undergoing imaging within three hours of stroke.
The ratio of sensitivity to specificity for FLAIR sequencing (sensitivity 90 percent, specificity 93 percent, Az equals .964) was more accurate than the DW (Az equals .749) and ADC (Az of.735) imaging ratios in differentiating patients imaged zero to three hours after stroke from those patients imaged more than three hours after stroke onset. The difference between the DW and ADC ratios was not significant. Employing FLAIR and DW imaging ratios together to determine time of stroke resulted in specificity and sensitivity of greater than 90 percent.
The researchers developed optimal cutoff values for differentiating the hyperacute and acute groups at 7 percent for FLAIR, meaning that in patients with sensitivity to specificity ratios below 7 percent, the values indicated that the patient had been imaged within three hours of stroke onset. This algorithm was the result of FLAIR's ratios (as well as the ratios for DW and ADC) exhibiting positive correlations with time. As time from stroke onset increased, the signal intensity seen on FLAIR images also increased.
Thus, patients presenting with normal intensity FLAIR images would be considered to have begun to experience stroke symptoms within three hours. Patients with definite hyperintensive FLAIR signals would be considered to have experienced stroke onset more than three hours before undergoing MRI. Finally, patients in the middle, with subtle hyperintensive FLAIR sequences, would be closer to the cutoff value and group placement would depend on their precise quantitate sensitivity to specificity ratio or area under the receiver operating characteristic curve (below 7 percent entailing three hours or less since stroke onset).
"Using MRI to determine the duration of a stroke could change the way patients with unknown stroke onset time are managed in emergency settings in two ways. First, among patients with unknown (or uncertain) stroke onset times, MRI can help identify those who are highly likely to be within the three-hour time window, when intravenous thrombolysis has been proven effective and is recommended by international guidelines. Second, by providing stroke teams with a reliable marker of stroke age, MRI could help ensure that patients with unknown stroke onset are managed as urgently as those with known stroke onset times," Oppenheim told Health Imaging News.
The authors pointed out several limitations to their study. The authors could not determine cutoff ratios for patients imaged within 4.5 hours of stroke onset, when thrombolysis could still be useful, because their sample only included 15 patients imaged within that time period, not a large enough sample for the determination of accurate cutoffs. Additionally, the researchers cautioned that their findings should not be extrapolated to MRI at field strengths greater than 1.5T, as different FLAIR and DW parameters would apply for time classification at higher intensities.
"Using the MRI data alone, our radiologists were able to predict with greater than 90 percent accuracy which patients had experienced stroke symptoms for longer than three hours. Such an MRI-based clock might increase the number of candidates for thrombolytic therapy, provided that such patients are transferred urgently to a hospital after stroke recognition. In other words," Oppenheim testified, "time is brain."