Texture-Based Quantitative Imaging Biomarkers for Tomographic Disease Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cancer biomarkers rely on morphological features and require image segmentation, which is operator-dependent, inaccurate, and limited by 2D image processing, failing to effectively measure disease progression and response to therapy without measurable tumor size changes.
Innovation Solution
The development of texture-based quantitative imaging biomarkers that extract features from tomographic images without segmentation, using Earth Movers Distance and other metrics to compute disease severity and progression, allowing for the identification and tracking of regions of interest over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If morphological features and image segmentation are used to measure tumor response, then tumor burden size can be computed, but the method becomes operator-dependent and inaccurate
Solution Approach 1:
The patent replaces manual image segmentation (mechanical/operator-dependent process) with automated texture analysis algorithms that compute quantitative biomarkers directly from tomographic images. This substitution eliminates operator dependency while maintaining measurement accuracy through objective computational methods.
Solution Approach 2:
The patent transitions from measuring morphological parameters (tumor size, shape) to measuring texture parameters (intensity variations, spatial patterns). This parameter change enables accurate tumor response assessment without requiring precise segmentation boundaries, as texture features can be extracted from entire regions of interest.
2Ease of operation
If 2D image processing techniques are used for lesion delineation, then image analysis can be performed, but spatial relationship measures between pixels are limited
Solution Approach 1:
The patent transitions from 2D slice-based analysis to 3D volumetric texture analysis. By processing the entire 3D volume and computing texture features across multiple slices, the system captures spatial relationships between pixels in three dimensions, significantly improving measurement precision while maintaining ease of operation through automated algorithms.
3Measurement precision
If tumor size changes are used as the basis for response evaluation, then morphological assessment can be performed, but disease progression cannot be measured when tumors show little or no shrinkage
Solution Approach 1:
The patent replaces morphological parameters (tumor size) with texture parameters (intensity distributions, spatial patterns) as the basis for response evaluation. This parameter change enables the assessment of disease progression through texture alterations even when tumor size remains stable, making the method applicable to cytostatic drugs that inhibit tumor growth without causing shrinkage.
4Difficulty of detecting and measuring
If image segmentation is performed to define regions of interest, then lesion identification can be achieved, but the process becomes highly operator-dependent and reduces precision
Solution Approach 1:
The patent extracts texture features directly from entire regions of interest without performing segmentation to isolate individual lesions. By computing texture biomarkers from the full ROI volume, the method eliminates segmentation-related errors and operator dependency while maintaining lesion identification capability through texture-based differentiation.
Data Source
AI summary
An image processing apparatus for computing a quantitative imaging biomarker (QIB) of disease severity from variations in texture-based features in a tomographic image, the apparatus including a first preprocessing module for normalizing the intensities in the tomographic image; a second identification module for identifying at least one organ of interest in the tomographic image; a third ROI selection module for identifying and selecting a plurality of target ROIs and reference ROIs representative respectively of abnormal and normal pathology in the organ(s) of interest; a fourth ROI assessment module for extracting a plurality of texture-based feature signatures from the target ROIs and the reference ROIs, wherein the feature signatures are generated from distributions of statistical attributes extracted from each ROI; a fifth biomarker assessment module for computing the distance between the target ROI signatures and the reference ROI signatures, wherein the biomarker of disease severity is a function of the distances between the target ROI signatures and the reference ROI signatures.