By Charles Bankhead

Diffusion-weighted MRI offers superior diagnostic capability for acute stroke, compared to screening CT scans of the head, results of a Harvard study suggest.

In one of the largest stroke imaging studies reported to date, MRI with diffusion-weighted imaging had a 97% sensitivity and 100% specificity for stroke detection. In contrast, an initial head CT scan had 40% sensitivity and 92% specificity for identifying acute stroke.

"From these data, we conclude that MRI with diffusion-weighted imaging is more sensitive than other modalities, and at least as specific, and offers better negative predictive value than the other imaging modalities," Harvard radiologist Dr. Mark Mullins said at the RSNA Scientific Sessions. "In head-to-head comparisons, diffusion-weighted MRI offers better accuracy than either screening CT or follow-up CT. Though none of the patients had conventional and diffusion-weighted MRI during the first six hours after presentation, we feel that diffusion-weighted imaging is probably more accurate than conventional MRI."

Mullins and his associates arrived at their conclusions after retrospectively reviewing medical records on 733 patients who presented to the emergency department and admitting with diagnoses of stroke. Stroke was confirmed in all cases by the clinical neurological diagnosis at discharge.

Of the entire group, 509 patients had screening CT of the head within the first six hours after presentation to the emergency department. Eight patients had a follow-up CT scan within the first six hours. Fourteen patients had conventional MRI head scans, and 122 patients had imaging evaluation by diffusion-weighted MRI.

The first screening CT resulted in 256 false negatives, seven false positives, 77 true negatives, and 169 true positives, which translated into 40% sensitivity, 92% specificity, 96% positive predictive value, and 23% negative predictive value. The eight follow-up CT scans resulted in a 71% sensitivity and 100% specificity, 100% positive predictive value, and 33% negative predictive value.

Conventional MRI for acute stroke had 58% sensitivity, 100% specificity, 100% positive predictive value, and 28% negative predictive value.

Diffusion-weighted MRI resulted in three false negatives, no false positives, 10 true negatives, and 109 true positives, which translated into 97% sensitivity, 100% specificity, 100% positive predictive value, and 77% negative predictive value.