Zone Percentage (ZP)

Zone Percentage is calculated as the number of voxels in the most frequent zone size divided by the total number of voxels in the ROI.

• Formula: ZP = (N_z_max / N_v) * 100 where N_z_max is the number of voxels in the largest zone size class and N_v is the total voxel count
• Range: 0 to 1 (or 0% to 100%)
• Interpretation: A high ZP indicates that a single zone size dominates the image, suggesting textural homogeneity
• Low ZP: Indicates a more even distribution across multiple zone sizes, reflecting heterogeneous texture

Zone Percentage is derived directly from the Gray-Level Size Zone Matrix (GLSZM), which quantifies connected regions of identical voxel intensity.

• The GLSZM counts zones independently of their rotational orientation, making it rotation-invariant
• ZP summarizes the matrix by identifying the dominant zone size class
• Unlike Zone Size Non-Uniformity (ZSNU), which measures variability across all zone sizes, ZP focuses specifically on the prevalence of the majority class
• Works in conjunction with Large Zone Emphasis (LZE) and Small Zone Emphasis (SZE) for comprehensive texture profiling

In oncology imaging, Zone Percentage serves as a biomarker for tissue architecture uniformity.

• High ZP in tumors: Often correlates with necrosis or homogeneous cellular packing, potentially indicating aggressive pathology
• Low ZP in tumors: May reflect spatial heterogeneity associated with treatment-resistant sub-regions
• Treatment response: A shift toward higher ZP post-therapy can indicate the development of uniform fibrotic tissue
• IBSI compliance: Must be calculated according to Image Biomarker Standardisation Initiative guidelines to ensure cross-institutional reproducibility

Zone Percentage is highly sensitive to preprocessing parameters that affect zone connectivity.

• Intensity Discretization: The number of gray-level bins directly impacts zone formation. Fewer bins create larger, more connected zones, artificially inflating ZP
• Voxel Resampling: Isotropic resampling is critical; anisotropic voxels distort zone size calculations
• Segmentation Accuracy: ZP is calculated only within the Region of Interest (ROI); segmentation errors at tumor boundaries propagate directly to the metric
• IBSI recommends: Fixed bin number (FBN) discretization with 32 or 64 bins for reproducible ZP values

Zone Percentage complements other GLSZM-derived features to provide a complete heterogeneity profile.

• vs. Zone Size Non-Uniformity (ZSNU): ZSNU measures the variability of zone size distribution; ZP measures the dominance of a single size
• vs. Large Zone Emphasis (LZE): LZE weights larger zones more heavily; ZP identifies whether any single size class dominates regardless of absolute size
• vs. GLCM Homogeneity: GLCM Homogeneity measures local intensity similarity; ZP measures regional size uniformity
• Combined use: Low ZP + High ZSNU indicates a tumor with highly variable, multi-scale textural patterns

Zone Percentage exhibits variable test-retest reliability depending on acquisition parameters.

• Intraclass Correlation Coefficient (ICC): Studies report ICC values ranging from 0.75 to 0.95 for ZP, depending on discretization settings
• Scanner variability: ZP is sensitive to reconstruction kernel and slice thickness differences across vendors
• ComBat Harmonization: Applying ComBat batch-effect correction can reduce inter-scanner ZP variance by up to 40%
• Delta-radiomics: Changes in ZP over time are more clinically meaningful than absolute values, as they cancel systematic scanner biases