|
| Sl. no | Feature type | Features |
|
| 1 | Texture features (33) | Gray-level co-occurrence matrix (GLCM)—inertia, energy, correlation, contrast, entropy, and homogeneity |
| Gabor wavelet |
| Statistical features—mean, median, kurtosis, and skewness |
| Local binary pattern (LBP)—textures spatial structure |
| Gray-level difference statistics (GLDS)—mean, entropy, contrast, angular second moment, and homogeneity |
| Fractal dimension texture analysis (FDTA) |
| Radial and angular sum of discrete Fourier transform for Fourier power spectrum |
| Neighborhood gray-tone difference matrix (NGTDM), given by strength, complexity, coarseness, and contrast |
| Absolute gradient—mean, variance |
|
| 2 | Shape features (5) | Sharpness |
| Complexity |
| Length irregularity |
| Aspect ratio and circularity |
|
| 3 | Histogram and correlogram features (10) | Gray-level histogram of segmented ROI of the carotid image—for 32 same measurements, bins were computed |
| Plaque histogram represents plaque characterization |
| Multiregion histogram—to check whether plaque outer region signifies disease progression |
| Grayscale median (GSM) derived from the histograms first-order statistics, and entropy represents echogenicity |
| Histogram of oriented gradient (HOG)—gradient magnitude and orientation |
| Correlogram—statistics and spatial distribution of the features |
| Texture and shape features—normalized; histogram and correlogram features—used without normalization |
|
| 4 | Morphology features (15) | Mean probability density functions (PDFs), mean cumulative distribution functions (CDFs) |
| Plaque power spectra for all the three intensity variations (low, medium, and high) |
| Shape, connectivity, and convexity |
| Plaque size, lipid core, and presence of intraplaque hemorrhage |
| Smooth lumen surface—no risk; rough lumen—leads to stroke |
| Plaque volume |
|
