Multi-Scale Classifier Fusion for Lesion Boundary Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for border detection in dermoscopy images are subjective and lack accuracy due to human interpretation variability, necessitating a more robust and automated approach for identifying lesion boundaries.
Innovation Solution
A multi-scale classification-based border detection method using supervised machine learning, where multiple classifiers trained on different image resolutions fuse their predictions to generate a probability map, which is then thresholded to create a binary mask for accurate lesion boundary identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple classifiers trained on different image resolutions are used, then border detection accuracy is improved, but device complexity increases
Solution Approach 1:
The system segments the border detection task into multiple independent classifiers, each trained on a specific resolution range. This segmentation allows each classifier to specialize in detecting borders at particular scales, improving overall accuracy while maintaining manageable complexity through modular design
Solution Approach 2:
The multi-scale classifier system provides universal border detection capability across varying image resolutions. Each classifier is designed to handle specific resolution ranges, and together they form a universal system that can accurately detect borders regardless of the input image scale, eliminating the need for separate specialized systems for different resolutions
2Measurement precision
If supervised machine learning with multiple classifiers is employed, then detection precision is improved, but loss of time increases due to multiple classification steps
Solution Approach 1:
The classifiers are pre-trained on diverse resolution ranges during the training phase, preparing them to handle various input scales. This preliminary action ensures that during actual border detection, the system can quickly apply pre-configured classifiers without requiring extensive real-time computation or adaptation, reducing processing time while maintaining high accuracy
3Reliability
If late fusion method combining multiple predicted borders is used, then reliability of border detection is improved, but device complexity increases
Solution Approach 1:
The late fusion process acts as an intermediary mechanism that combines the predictions from multiple classifiers. By introducing this systematic fusion step, the system reliably integrates results from different resolution-specific classifiers, producing a consolidated and reliable border detection outcome while managing complexity through a structured combination process
Data Source
AI summary
A robust segmentation technique based on multi-layer classification technique to identify the lesion boundary is described. The inventors have discovered a technique based on training several classifiers such that to classify each pixel as lesion versus normal Each classifier is trained on a specific range of image resolutions. Then, for a new test image, the trained classifiers are applied on the image. Then by fusing the prediction results in pixel level a probability map is generated. In the next step, a thresholding method is applied to convert the probability map to a binary mask, which determines a mole border.