Short Run Emphasis (SRE)

SRE is calculated by dividing the sum of all run lengths divided by their squared values by the total number of runs in the GLRLM. The formula is:

SRE = Σ [P(i,j) / j²] / Σ P(i,j)

• P(i,j): The number of runs with gray level i and run length j
• j² in denominator: Heavily penalizes long runs, giving disproportionate weight to short runs
• Normalization: Dividing by total runs ensures the metric is independent of ROI size
• Value range: Typically between 0 and 1, where higher values indicate finer textures

SRE serves as a quantitative proxy for spatial frequency within a region of interest:

• High SRE (>0.7): Indicates a predominance of short runs (1-2 pixels), corresponding to fine, heterogeneous textures with rapid intensity fluctuations
• Low SRE (<0.3): Reflects longer runs of uniform intensity, characteristic of coarse, homogeneous textures with gradual intensity transitions
• Clinical correlation: Fine textures often correspond to highly cellular regions, while coarse textures may indicate necrosis or cystic degeneration

SRE is inversely related to Long Run Emphasis (LRE), its complementary GLRLM feature:

• Inverse weighting: SRE uses 1/j² while LRE uses j², creating a reciprocal relationship
• Run length spectrum: Together, SRE and LRE characterize the full distribution of texture granularity
• Run Percentage (RP): Another related metric measuring the fraction of runs at any length
• Gray-Level Non-Uniformity (GLN): Measures the similarity of gray-level values throughout runs, independent of run length

A balanced SRE/LRE ratio often indicates mixed tissue architecture with both fine and coarse elements.

SRE has demonstrated prognostic value across multiple cancer types:

• Lung cancer: High SRE in CT scans correlates with lepidic growth patterns and better survival in adenocarcinoma
• Glioblastoma: Elevated SRE on MRI differentiates pseudoprogression from true tumor recurrence
• Breast cancer: Low SRE on mammography associates with invasive ductal carcinoma versus benign lesions
• Colorectal cancer: SRE changes during chemotherapy predict pathological complete response

SRE captures intratumoral heterogeneity that may be invisible to the naked eye.

SRE values are sensitive to image acquisition and preprocessing parameters:

• Intensity discretization: The number of gray-level bins directly impacts run length counts. IBSI recommends 32-64 bins for consistent SRE calculation
• Pixel/voxel size: Anisotropic resolution creates directional bias in run length measurement
• Interpolation effects: Resampling to isotropic voxels alters the spatial relationships between pixels
• Noise sensitivity: Image noise creates artificial short runs, inflating SRE values

ComBat harmonization is often applied to mitigate scanner-specific SRE variations in multi-center studies.

The parent matrix from which Short Run Emphasis (SRE) is derived. The GLRLM quantifies texture by counting consecutive, collinear pixels that share the same gray-level intensity. It captures directional roughness by evaluating runs in 13 directions in 3D space. SRE specifically measures the distribution of these runs, with a high value indicating a predominance of short runs and thus a fine, heterogeneous texture.

A second-order statistical method that analyzes the spatial relationship of pixel pairs. Unlike the GLRLM's focus on runs, the GLCM calculates how often a pixel with intensity i occurs adjacent to a pixel with intensity j. Key derived features include Contrast, Homogeneity, and Correlation. While SRE measures run-length fineness, GLCM features describe the overall smoothness and uniformity of local intensity transitions.

Quantifies the size of homogeneous, connected regions of identical voxel intensity, independent of their rotational orientation. A key feature is Zone Percentage (ZP), which measures the homogeneity of zone sizes. While SRE focuses on the length of 1D runs, GLSZM analyzes the area/volume of 2D/3D zones, making it highly effective for characterizing non-linear, clumped heterogeneity in tumors.

Measures the average absolute difference between a voxel's intensity and the mean intensity of its surrounding neighbors within a Chebyshev distance. Features like Coarseness, Contrast, and Busyness describe the spatial rate of intensity change. A high SRE (fine texture) often correlates with a high Busyness value in the NGTDM, as both indicate rapid spatial fluctuations in pixel intensity.

The direct mathematical complement to Short Run Emphasis. LRE is a GLRLM-derived metric that quantifies the distribution of long consecutive runs of identical pixel intensities. A high LRE value indicates a coarse, homogeneous texture with large structural patterns. Comparing SRE and LRE provides a powerful, scale-invariant summary of a tissue's architectural complexity.

A critical preprocessing step for all texture matrices, including the GLRLM used to calculate SRE. This process bins continuous voxel intensity values into a finite number of discrete gray levels (e.g., 8, 16, 32, 64 bins). The choice of bin width directly impacts SRE values; a smaller number of bins can merge distinct intensities, artificially creating longer runs and lowering SRE, making standardization via IBSI guidelines essential.