Neighborhood Gray-Tone Difference Matrix (NGTDM)

A measure of the spatial rate of intensity change. Coarseness quantifies the granularity of the texture by summing the product of each gray-level probability and its corresponding NGTDM entry.

• High Coarseness: Indicates a uniform, slowly changing texture with large, homogeneous regions. Spatial variations are low.
• Low Coarseness: Indicates a fine, rapidly changing texture with high spatial frequency and small, distinct primitives.
• Clinical Relevance: In oncology, higher coarseness often correlates with more homogeneous tumor tissue, while lower values suggest greater microstructural heterogeneity.

Quantifies the intensity difference between neighboring regions and the dynamic range of the gray levels. Contrast is high when the image has sharp intensity transitions and a wide range of gray values.

• Calculation: It weights the squared difference between gray-level pairs by their average probability, emphasizing large intensity excursions.
• High Contrast: Suggests a heterogeneous texture with distinct, sharply delineated boundaries between tissue compartments.
• Low Contrast: Suggests a blurred, smooth texture with minimal intensity variation between adjacent regions.
• Example: A CT scan with a necrotic core adjacent to a highly vascularized rim will exhibit high contrast.

Measures the spatial frequency of intensity changes. Busyness is high when intensity changes rapidly from one pixel to its neighborhood, indicating a 'busy' texture.

• Mechanism: It computes the sum of the product of gray-level probabilities and their NGTDM values, normalized by the absolute differences between gray levels.
• High Busyness: Indicates a fine, chaotic texture with rapid, high-frequency intensity fluctuations.
• Low Busyness: Indicates a coarse, slowly varying texture.
• Distinction from Coarseness: While coarseness measures the magnitude of the difference, busyness emphasizes the rate of change between adjacent intensity levels.

A measure of the information content and the number of primitive components in the texture. Complexity is high when there are many distinct gray-level patches and rapid spatial changes.

• Calculation: It sums the absolute differences between the probabilities of gray-level pairs, weighted by the sum of their NGTDM values.
• High Complexity: Indicates a non-uniform, intricate texture with many different intensity primitives and irregular borders.
• Low Complexity: Indicates a simple, regular texture with few distinct gray-level components.
• Clinical Context: High complexity in a tumor ROI often indicates a disorganized, aggressive growth pattern with mixed cellularity and necrosis.

A measure of the overall emphasis on coarse, high-contrast primitives relative to fine, low-contrast ones. Strength is high when the image contains well-defined, easily distinguishable coarse structures.

• Mechanism: It is computed as the ratio of the sum of the product of probabilities and squared gray-level differences to the sum of the NGTDM values.
• High Strength: Indicates that the texture is dominated by bold, coarse, and visually prominent structures.
• Low Strength: Indicates that fine, low-contrast textures dominate the image.
• Interpretation: Strength helps differentiate between images where coarse features are visually dominant versus those where they are present but not the primary textural signal.

The signature feature derived from NGTDM, quantifying the spatial rate of intensity change. High coarseness indicates large, uniform regions; low coarseness indicates fine, rapidly changing textures.

• Formula: Coarseness = 1 / (sum of NGTDM elements + small epsilon)
• Inversely related to the sum of neighborhood differences
• Clinical relevance: High coarseness in tumors often correlates with necrosis
• Highly sensitive to discretization parameters
• Often combined with contrast and busyness for multi-feature texture profiling