Sphericity

Sphericity is calculated as the ratio of the surface area of a sphere with the same volume as the tumor to the actual surface area of the tumor. The formula is:

• Sphericity = (∏^(1/3) * (6V)^(2/3)) / A
• Where V is the volume and A is the surface area of the segmented region of interest
• Values range from 0 to 1, where 1 represents a perfect sphere
• The metric is inherently rotationally invariant and does not depend on the absolute size of the lesion

Sphericity serves as a surrogate marker for tumor aggressiveness and growth patterns:

• High Sphericity (>0.8): Often associated with benign, well-circumscribed lesions or slow-growing tumors that expand uniformly
• Low Sphericity (<0.6): Typically indicates malignant, infiltrative processes with irregular, spiculated margins invading surrounding tissue
• Serial Monitoring: A decreasing sphericity value over time can signal a transition to a more aggressive phenotype or treatment resistance

The Image Biomarker Standardisation Initiative (IBSI) provides a consensus definition to ensure reproducibility across platforms:

• The feature is defined under the morphological/shape category in the IBSI reference manual
• Requires a 3D binary mask of the segmented tumor volume
• Mesh-based surface area calculation using marching cubes triangulation is recommended to avoid pixelation artifacts
• IBSI benchmarks provide reference values for digital phantoms to validate algorithm implementations

Accurate sphericity calculation depends heavily on preprocessing steps:

• Segmentation Quality: Errors in tumor boundary delineation, especially at spiculated edges, propagate directly into surface area estimates
• Voxel Resolution: Anisotropic voxel dimensions require resampling to isotropic grids to prevent geometric distortion
• Surface Smoothing: Excessive mesh smoothing can artificially inflate sphericity values by removing clinically relevant irregular margins
• Partial Volume Effect: Small tumors occupying only a few voxels yield unreliable shape metrics due to discretization noise

Sphericity belongs to a family of related 3D morphological descriptors, each capturing a distinct aspect of tumor geometry:

• Compactness: Measures how efficiently the volume is packed relative to a sphere (V / (√(A³)))
• Surface-to-Volume Ratio: Quantifies the irregularity of the tumor boundary independently of sphericity's normalization
• Elongation: Captures the major-to-minor axis ratio, which sphericity does not explicitly measure
• Asphericity: The complementary metric (1 - Sphericity), directly quantifying the deviation from a spherical shape

Sphericity has demonstrated independent prognostic significance across multiple cancer types:

• Non-Small Cell Lung Cancer: Lower sphericity on baseline CT correlates with reduced overall survival and higher recurrence rates
• Glioblastoma: Spherical tumors with necrotic cores exhibit distinct genetic profiles compared to irregularly shaped infiltrative lesions
• Breast Cancer: Sphericity of the primary tumor on MRI is a predictor of pathological complete response to neoadjuvant chemotherapy
• Head and Neck Squamous Cell Carcinoma: Shape features outperform volume alone in predicting locoregional control

A dimensionless measure of how efficiently a tumor's volume is packed relative to its surface area. Compactness compares the shape to a sphere, where a value of 1 indicates perfect spherical packing. Irregular, infiltrative tumors exhibit lower compactness values.

• Formula: ( \text{Compactness} = \frac{36\pi V^2}{A^3} )
• Highly sensitive to spiculations and surface irregularities
• Often used alongside sphericity in radiomic signatures for glioblastoma grading

Captures the degree to which a tumor deviates from a sphere by stretching along a single axis. Elongation is calculated from the ratio of the major to minor principal axes of the region of interest.

• Derived from the eigenvalues of the covariance matrix
• A perfect sphere has an elongation of 1.0
• Critical for distinguishing spindle-cell lesions from round-cell tumors in sarcoma classification

Measures the extent to which a tumor is compressed along one axis, approximating a disc or planar structure. Flatness is computed from the ratio of the smallest to largest principal component eigenvalues.

• Values near 0 indicate extreme flatness
• Values near 1 indicate a spherical or cylindrical shape
• Useful in characterizing meningiomas and plaque-like lesions adjacent to anatomical boundaries

A fundamental morphological metric quantifying the relationship between a tumor's boundary area and its enclosed volume. Higher ratios indicate more irregular, infiltrative margins with greater surface complexity.

• Directly correlates with tumor invasiveness in breast cancer
• Measured in mm⁻¹
• Highly dependent on segmentation accuracy and voxel resolution

The largest pairwise Euclidean distance between any two surface voxels in the three-dimensional region of interest. This RECIST-correlated measurement represents the longest axis of the tumor.

• Directly comparable to manual radiological caliper measurements
• Used as a reference standard for validating automated segmentation tools
• Forms the basis for volumetric response criteria in oncology trials

Quantifies the asymmetry of a tumor by comparing its actual surface area to the surface area of a sphere with equivalent volume. Spherical disproportion increases with surface irregularity and non-spherical morphology.

• Formula: ( \text{SD} = \frac{A}{\sqrt[3]{36\pi V^2}} )
• A perfect sphere yields a value of 1.0
• Sensitive to lobulated contours in renal cell carcinoma grading