Image Biomarker Standardisation Initiative (IBSI)

Establishes a controlled vocabulary for radiomic features, eliminating ambiguity across research groups. Each feature receives a unique, canonical identifier.

• Goal: Ensure that 'Entropy' calculated in one lab matches the definition in another.
• Mechanism: Maps common synonyms and vendor-specific names to a single IBSI label.
• Impact: Enables direct comparison of results in multi-center clinical trials.

Defines the exact mathematical equations and algorithmic steps for computing each feature, resolving implementation discrepancies.

• Core Principle: Provides a 'ground truth' formula to test software against.
• Example: Specifies the precise binning strategy and aggregation method for a Gray-Level Co-occurrence Matrix (GLCM).
• Outcome: Transforms proprietary 'black box' calculations into transparent, auditable processes.

Provides a publicly available set of synthetic images and corresponding validated feature values to verify software implementations.

• Function: A 'unit test' for radiomics software.
• Process: Users run their pipeline on the IBSI phantom and compare output values to the gold-standard reference.
• Validation: Passing the benchmark confirms that a tool is IBSI-compliant, a critical step for regulatory acceptance.

Specifies the minimum metadata and processing parameters that must be reported to ensure a study is reproducible.

• Key Parameters: Includes intensity discretization method, bin width, and interpolation strategy.
• Purpose: Prevents the 'file drawer problem' where features are non-reproducible due to undocumented preprocessing.
• Adoption: Aligns with broader imaging biomarker reporting checklists for journal submissions.

Addresses the challenge of feature stability across different scanner vendors and acquisition protocols.

• Strategy: Provides a framework for applying ComBat harmonization and other batch-effect correction techniques.
• Focus: Distinguishes between biological variance and technical noise introduced by CT/PET/MR scanner variability.
• Result: Facilitates robust multi-center data pooling without losing diagnostic signal.

A batch-effect correction method adapted from genomics to address the primary barrier to multi-center radiomic validation: scanner variability.

• Removes non-biological technical variance introduced by manufacturer, acquisition protocol, or reconstruction kernel
• Preserves biological covariates like tumor grade or patient age during correction
• Essential for achieving IBSI-compliant reproducibility across heterogeneous datasets

The critical preprocessing step of binning continuous Hounsfield Units or signal intensities into a finite number of gray levels. IBSI mandates specific discretization strategies to ensure texture matrix reproducibility.

• Fixed Bin Number (FBN): Divides the intensity range into a constant number of bins
• Fixed Bin Width (FBW): Uses a constant bin size in original intensity units
• Directly impacts GLCM, GLRLM, and GLSZM feature values

The primary statistical metric for quantifying test-retest reliability and inter-observer agreement of radiomic features.

• ICC > 0.75 typically defines a feature as stable and suitable for model inclusion
• Used to filter out features vulnerable to segmentation variability before LASSO or mRMR selection
• IBSI reference manuals recommend ICC analysis as a mandatory robustness checkpoint

A composite biomarker constructed from a selected panel of IBSI-compliant features combined via a mathematical model to predict a specific clinical endpoint.

• Built using feature selection algorithms like LASSO or mRMR to prevent overfitting
• Validated on independent external cohorts to demonstrate generalizability
• Represents the translational endpoint of the IBSI standardization pipeline

The extraction and analysis of changes in quantitative imaging features over time or across treatment cycles. IBSI standardization enables the reliable comparison of feature values between longitudinal scans.

• Quantifies early therapeutic response before anatomical shrinkage occurs
• Requires precise image registration and consistent acquisition parameters
• Identifies sub-regions of evolving tumor heterogeneity via habitat imaging