Article Figures & Data
Figures
- Fig 1.
These images represent 9 contiguous pixels representing the same structure that spans these pixels. In an ideal noise-free image (A), these 9 pixels would have identical intensities. The presence of noise that invariably accompanies digital images causes intensities of some of these pixels to be higher or lower than expected (B). Modifying the intensities of pixels (asterisk) on the basis of the preponderance of intensities (plus sign) in their neighboring pixels can mitigate the effect of noise (C).
- Fig 2.
Axial head CT image across the centrum semiovale before (baseline image) and after (enhanced image) processing with CIE, with equivalent ROIs used to measure gray and white matter intensities. These measurements were used to calculate the conspicuity of gray matter and CNR between gray and white matter.
- Fig 3.
Boxplots showing the distribution of gray matter conspicuity (A) and CNR between gray and white matter (B) in 60 paired head CT images with normal findings before (baseline) and after (enhanced) processing with CIE. Both of these quantitative measures of gray-white differentiation demonstrated significant improvement following image processing (P < .001 for conspicuity; P = .015 for CNR).
- Fig 4.
Clustered columns showing the distribution of qualitative GWD ratings assigned by 2 blinded radiologists to baseline and enhanced images on a 5-point Likert scale ranging from 1 (imperceptible GWD) to 5 (very easily perceptible GWD). The vertical axis shows the percentage of all cases assigned a given GWD rating. A favorable shift toward higher GWD ratings was seen with image processing for both radiologists (P < .01).
- Fig 5.
Axial head CT image across the insula obtained within 3 hours of stroke onset before (A) and after (B) processing with CIE. Note that improved gray-white differentiation following image processing makes it easier to perceive the loss of normal gray matter density in the insula (arrow), corresponding to the infarction proved on subsequent DWI (C).
