Handheld Skin Lesion Classification Using Segmented Texture Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies lack effective algorithms for the detection, analysis, and classification of skin and ocular diseases, as well as plant diseases, that can operate on smart handheld devices with limited memory and computational speed, necessitating improved image processing and texture analysis methods.
Innovation Solution
A portable imaging system equipped with a digital camera, display, memory, processor, and network connection, utilizing a library of algorithms for image segmentation, feature extraction, and classification, including support vector machines, to identify and classify objects of interest on human or plant bodies in real time, even on devices with limited resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex image processing and texture analysis algorithms are used for accurate skin lesion detection and classification, then diagnostic accuracy is improved, but computational resource requirements exceed the capabilities of handheld devices with limited memory and processing speed
Solution Approach 1:
The patent segments the skin lesion image into multiple regions of interest (such as pigment network, globules, streaks, and other structures) and applies specific texture analysis algorithms to each segment. This allows complex diagnostic tasks to be divided into manageable parts that can be processed within the computational constraints of handheld devices while maintaining high diagnostic accuracy through focused analysis of each lesion component
Solution Approach 2:
The patent extracts specific texture features and structural characteristics from the skin lesion images, isolating the most diagnostically relevant information (such as pigment distribution patterns, vascular structures, and surface texture) from the full image data. This extraction process reduces the computational burden by focusing only on critical features needed for accurate classification, enabling complex analysis on resource-limited handheld devices
2Adaptability or versatility
If comprehensive image analysis algorithms are implemented for detecting multiple skin conditions and plant diseases, then diagnostic capability is improved, but the algorithms cannot run on devices with limited memory and computational speed
Solution Approach 1:
The patent applies different texture analysis algorithms and processing techniques to different regions and types of lesions based on their specific characteristics. For example, different algorithms are used for melanoma detection versus plant disease identification, and different texture metrics are applied to different lesion zones. This localized approach enables comprehensive multi-disease diagnostic capability while optimizing processing speed by applying only the necessary analysis to each specific case
Solution Approach 2:
The patent implements a tiered analysis approach where a preliminary quick assessment is performed first using simplified algorithms, followed by more comprehensive analysis only for lesions that require further evaluation. This partial action strategy enables the system to handle multiple disease types with varying levels of diagnostic depth, maintaining both versatility and processing efficiency by avoiding exhaustive analysis of all images
Data Source
AI summary
Provided herein are digital-implemented methods for performing simultaneous analyses on an object on the skin of an animal body, for example, a human, to classify the object as a skin cancer, an ulcer or neither. The analyses are performed simultaneously on a hand-held imaging device.