Lesion Segmentation via Layered Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional medical imaging techniques often treat tumors as single entities, limiting the information available on their structural complexity and growth patterns, which hampers accurate classification and diagnosis.
Innovation Solution
The method involves segmenting tumors into multiple layers using morphological erosion and subtraction of binary image masks, allowing for the extraction of features from each layer, which are then used for improved classification and characterization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If tumors are treated as single entities using conventional medical imaging techniques, then the analysis process is simple, but the information on structural complexity and growth patterns is limited
Solution Approach 1:
The patent applies segmentation by dividing the tumor into multiple concentric layers (outer layer, middle layer, inner layer) based on binary image masks. This allows extraction of features from each layer separately, capturing structural complexity and growth patterns that would be lost if the tumor were analyzed as a single entity. The segmentation process creates layered regions that represent different zones of the tumor, enabling detailed analysis of heterogeneity.
2Measurement precision
If tumors are segmented into multiple layers for detailed analysis, then classification accuracy is improved, but the processing complexity increases
Solution Approach 1:
The patent applies local quality by extracting features specifically from each tumor layer (outer, middle, inner) rather than treating the entire tumor uniformly. Different feature sets are computed for different layers, allowing the analysis to capture local characteristics and heterogeneity within the tumor. This localized feature extraction improves classification precision by accounting for regional variations in tumor properties.
Solution Approach 2:
The patent introduces a new dimension of analysis by adding the layer dimension to the traditional single-tumor analysis. Instead of analyzing one tumor as a whole, the system analyzes multiple layers (outer, middle, inner) as separate entities, each contributing its own feature set. This dimensional expansion from single-entity to multi-layer analysis enables more sophisticated classification while systematically managing the increased complexity through structured feature extraction.
Data Source
AI summary
A method comprising using at least one hardware processor for: receiving a digital medical image and a binary image mask, wherein the binary image mask depicts a segmentation of a lesion in the digital medical image; computing a plurality of layers of the lesion; for each of the plurality of layers of the lesion, extracting layer features; and sending the extracted layer features to a lesion classifier.