Skin Lesion Detection Output Moderation Using User-Specific Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing skin lesion detection models suffer from low accuracy, leading to inaccurate detection results and diminished user trust due to factors like environment lighting, camera quality, and erroneous detection of skin features.
Innovation Solution
A system that modifies the output of skin lesion detection models using user-specific data, such as known skin features, attention areas, and preferences, to improve accuracy and relevance for individual users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional camera technology is used for skin lesion detection, then the detection process is simple and accessible, but the accuracy of detection results is low
Solution Approach 1:
The patent introduces user data as an intermediary element that mediates between the detection model and the final output. User data includes information about the user's skin features, preferences, and historical data, which are used to modify and refine the detection results. This intermediary layer enables the system to adapt to individual users without requiring complex hardware modifications, thereby improving detection accuracy while maintaining relative system simplicity.
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting detection parameters based on user-specific data. The system modifies detection thresholds, prioritization levels, and result interpretations according to the user's skin type, known features, and preferences. This allows the same detection model to achieve high accuracy across different users by adapting parameters to individual characteristics rather than requiring complex model variations.
2Measurement precision
If detection models are trained on general data, then the model is simple and easy to deploy, but the results are inaccurate for individual users
Solution Approach 1:
The patent implements preliminary action by collecting and storing user data before the actual detection process. User data including skin feature information, historical detection results, and preferences are gathered in advance and stored in a database. This preliminary data collection enables the system to quickly retrieve and apply user-specific information during detection, achieving high individual accuracy without requiring complex real-time adaptation mechanisms.
Solution Approach 2:
The patent employs feedback mechanisms where the system continuously learns from user interactions and detection results. User feedback about detected lesions, correction of false positives, and preference indications are fed back into the system to refine future detections. This feedback loop enables the model to adapt to individual users over time, improving accuracy for each user without requiring complex initial training for every possible user scenario.
3Reliability
If the detection model outputs raw results, then the processing is fast and simple, but erroneous detections occur reducing user trust
Solution Approach 1:
The patent uses user data as an intermediary filtering layer between the raw detection model output and the final presented results. The system compares detected lesions against user-specific skin feature data and known features to filter out erroneous detections. This intermediary verification process significantly improves reliability and user trust by eliminating false positives while adding minimal processing complexity through efficient data comparison.
Solution Approach 2:
The patent replaces complex mechanical verification processes with data-driven filtering mechanisms. Instead of using complex image processing or multiple detection models, the system substitutes these with efficient data comparison operations between detection results and stored user data. This substitution maintains high reliability through intelligent filtering while reducing processing complexity through optimized data retrieval and comparison algorithms.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
According to an aspect, there is provided an apparatus (10) for moderating a skin lesion detection model (30). The apparatus (10) comprising a processor (12) and a memory (11). The processor (12) is configured to: receive user data (13) associated with a user; receive an output of the skin lesion detection model (14); modify the output of the skin lesion detection model (14) in accordance with the user data (13) to provide a modified output (15) indicative of skin lesion detection for the user; and output the modified output (15).