Vikas Pandey, MD
Stroke imaging is a very sophisticated area that has been constantly evolving leaving the burden on neurologists to be able to maintain the ability for proper interpretation of different imaging modalities and sequences. Series such as DWI and FLAIR are cornerstones to stroke diagnosis and ischemic lesion aging, but whether a lesion that is seen on DWI is also present on FLAIR is a matter that, many times, is up for debate between treating team members. To help solve the argument, the authors utilized a method similar to what was done to televisions back in the 1950s, they added color! They used color-coded FLAIR imaging with the idea that this would improve inter-rater agreement on determination if lesions were present on FLAIR or not for improved ischemic lesion aging.
The authors used retrospective case design to view the records of 113 ischemic stroke patients who were admitted to Asan Medical Center in Korea within 24 hours of clear onset of symptoms. The patients received a 1.5 Tesla MRI scan and the DWI slice with the largest lesion and the slice of corresponding FLAIR image were prepared (both the conventional FLAIR, as well as 256-color coded voxel imaging using auto-contrast to assign color intensity). These three slices were prepared and administered to two groups, well-trained stroke neurologists and neurologists is residency training. To avoid learning effect, they presented the images with DWI with the color-coded image first, then one week later with the conventional FLAIR image. These were categorized into obvious, subtle and negative (which would be positive for DWI-FLAIR mismatch). The results showed that the average NIHSS was 4, and the patients received the imaging approximately 7 hours on-average from symptom-onset. 13.3% had received IV tPA before and during MRI. The inter-rater agreement of DWI-FLAIR mismatch was kappa= 0.538 with conventional images and kappa=0.754 with color and the p-value for difference in percent agreement was p=0.004. The categorization of subtle change was higher in conventional than with color-coded images (38.1% vs 25.7%, p=0.01). 34 patients out of the 113 had discrepant ratings with conventional images however this discrepancy was resolved in 28/34 patients (82.4%) with the color-coded. These 28 patients had a higher amount of cardioembolic strokes and cortical lesion strokes. The positive predictive value of DWI-FLAIR mismatch predicting lesion within 4.5 hours was 85.3% and 71.9% with conventional FLAIR images, and 95.7% and 82.1% with color-coded FLAIR images for the experienced and resident groups, respectively.
The data appears to be most applicable for cardioembolic and cortical stroke lesions as it is these that seem to be not detected in conventional flair imaging (less inter-rater agreement) which improved with color-coded imaging. The idea that occasionally some of these lesions may be missed on conventional FLAIR is frightening given the concept that these usually signify older lesions that may increase hemorrhagic risk if thrombolysis is given. The data also showed that those lesions rated “subtle” decreased with color-coded imaging because the color contrast made the lesions more obvious given that the heterogenous nature of infarcted tissue comes out more with color coding. The improved positive predictive value of being within 4.5 hours of time of onset has therapeutic implications as this modality can be used in those with unknown time of onset as well as the wake-up stroke population. The use of color and resolution has improved quality of things such as high definition televisions and video gaming systems and no doubt stroke imaging modalities will improve in much the same manner with similar improvements in color, resolution and imaging strength. In thirty years, perhaps clinicians will look back at how primitive we were with things such as DWI and FLAIR and be shocked with how we hadn’t come up with things like 30 Tesla HD 1080p No-Radiation 4-D Super Simultaneous Thrombolysis/CT/MRI/Perfusion/Angiogram/Facebook-Uploading Scanners that they may be using at that time.
@DrVikasNeuro
