Texture Analysis

A second-order statistical method that quantifies texture by calculating how often pairs of pixels with specific values occur in a defined spatial relationship.

• Computes joint probability distributions of pixel pairs at specified distances and angles (0°, 45°, 90°, 135°)
• Derives key Haralick features including contrast, correlation, energy, and homogeneity
• Captures directional heterogeneity patterns invisible to first-order histogram analysis
• Example: High contrast values in a liver lesion GLCM indicate malignant heterogeneity, while low contrast suggests benign tissue

A texture matrix that counts the number of consecutive pixels with the same gray-level value in a specific direction to capture structural coarseness.

• Measures run-length emphasis — long runs indicate coarse, homogeneous textures; short runs indicate fine, heterogeneous textures
• Produces features like Short Run Emphasis (SRE) and Long Run Emphasis (LRE)
• Computed across multiple directions to detect anisotropic tissue patterns
• Example: High run-length non-uniformity in glioblastoma MRI correlates with aggressive tumor infiltration

A texture matrix that quantifies the size of connected regions of identical gray-level values, independent of their directional orientation.

• Rotation-invariant by design — captures texture coarseness regardless of structure orientation
• Key features include Small Zone Emphasis (SZE) and Large Zone Emphasis (LZE)
• Particularly effective for characterizing necrotic cores and cystic regions in tumors
• Example: High zone-size variance in lung nodule CT scans strongly associates with malignant classification in early-stage adenocarcinoma

A texture matrix that quantifies the difference between a pixel's gray value and the average gray value of its surrounding neighbors within a defined distance.

• Captures local intensity variation — high values indicate rapid transitions between tissue types
• Produces five core features: coarseness, contrast, busyness, complexity, and strength
• Sensitive to boundary transitions between anatomical structures
• Example: Elevated NGTDM busyness in breast DCE-MRI reflects angiogenic heterogeneity characteristic of aggressive tumor subtypes

A mathematical decomposition technique that filters an image into different frequency sub-bands to extract multi-scale texture features not visible in the spatial domain.

• Applies high-pass (H) and low-pass (L) filters across all three dimensions, producing 8 decomposition bands (LLL, LLH, LHL, LHH, HLL, HLH, HHL, HHH)
• Texture matrices computed on each sub-band capture scale-specific heterogeneity patterns
• Reveals micro-architectural details obscured in original resolution
• Example: Wavelet-filtered GLCM features from HHH sub-bands capture fine trabecular bone texture predictive of osteoporotic fracture risk

An edge-detection filter that applies Gaussian smoothing before computing the Laplacian to highlight regions of rapid intensity change at various scales.

• The sigma parameter controls the scale of analysis — small sigma detects fine edges, large sigma detects coarse boundaries
• Acts as a band-pass filter in spatial frequency domain, suppressing both noise and large-scale intensity variations
• Texture features extracted from LoG-filtered images capture multi-scale structural transitions
• Example: LoG-filtered GLSZM features at sigma=2.0mm in prostate MRI differentiate Gleason grade groups with higher accuracy than unfiltered features

Texture analysis distinguishes malignant from benign lesions by quantifying the chaotic micro-architecture invisible to the human eye. Gray-Level Co-occurrence Matrix (GLCM) features like contrast and homogeneity correlate directly with tumor cellularity and nuclear atypia. In gliomas, texture heterogeneity derived from T2-weighted FLAIR MRI differentiates low-grade from high-grade tumors without biopsy, while in breast imaging, entropy and kurtosis values from dynamic contrast-enhanced MRI improve the positive predictive value of BI-RADS assessments.

Delta-radiomics captures the temporal evolution of texture features during therapy, often revealing response weeks before gross anatomical changes manifest. Key applications include:

• Immunotherapy: A decrease in tumor entropy on CT scans correlates with pseudoprogression, distinguishing true progression from inflammatory flare
• Chemoradiation: Rising GLCM homogeneity in rectal cancer MRI after neoadjuvant therapy predicts pathologic complete response
• Anti-angiogenic therapy: Changes in Gray-Level Run Length Matrix (GLRLM) run emphasis reflect vascular normalization before tumor shrinkage occurs

Texture-based radiomic signatures serve as independent prognostic biomarkers, often outperforming traditional staging systems. In non-small cell lung cancer, a composite of GLCM correlation and Gray-Level Size Zone Matrix (GLSZM) zone percentage extracted from pre-treatment CT predicts overall survival with a concordance index exceeding 0.70. In head and neck squamous cell carcinoma, coarseness from the Neighborhood Gray-Tone Difference Matrix (NGTDM) stratifies patients into distinct risk groups, enabling treatment intensification for high-risk cohorts.

Texture analysis bridges the gap between imaging phenotypes and underlying molecular biology, a field termed radiogenomics. Specific texture patterns correlate with:

• EGFR mutations in lung adenocarcinoma: Lower GLCM entropy and higher homogeneity on CT
• IDH1 mutation status in gliomas: Increased GLSZM size-zone variability on T2-weighted MRI
• Microsatellite instability in colorectal cancer: Higher skewness and kurtosis on contrast-enhanced CT These non-invasive surrogates guide targeted therapy selection when biopsy tissue is insufficient.

Beyond oncology, texture analysis characterizes atherosclerotic plaque vulnerability. GLCM features extracted from coronary CT angiography differentiate lipid-rich necrotic cores from fibrous tissue, identifying high-risk plaques prone to rupture. In carotid ultrasound, run-length non-uniformity from GLRLM quantifies plaque heterogeneity, independently predicting cerebrovascular events. Texture-derived entropy of myocardial tissue on cardiac MRI also detects diffuse fibrosis in cardiomyopathies before functional impairment becomes apparent.

Texture analysis quantifies the subtle microstructural brain changes that precede volumetric atrophy. In Alzheimer's disease, hippocampal texture entropy on high-resolution T1-weighted MRI discriminates mild cognitive impairment from healthy controls with higher sensitivity than volumetry alone. In multiple sclerosis, GLCM inverse difference moment of normal-appearing white matter detects occult pathology invisible on conventional FLAIR sequences, providing an imaging biomarker for disease progression in clinical trials.