Additive AI Models for Veterinary Radiology Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI systems in veterinary radiology face inefficiencies due to the need for manual identification of body regions in images, leading to errors and inefficiencies, and the conventional approach of replacing AI models assumes 'noise' is not valuable, resulting in lost data when retraining.
Innovation Solution
A method involving additive AI results from multiple classifiers, where a first AI classifier's results are compared to a derivative classifier's results, and both are used to generate a synthesis processor output, incorporating both old and new models for improved performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual identification of body regions is performed, then processing accuracy is improved, but productivity deteriorates due to extensive manual effort and time consumption
Solution Approach 1:
The system enables self-service by implementing automated body region identification using AI processors that independently analyze radiologic images without requiring manual identification. The AI processors automatically detect and segment body regions, then route images to appropriate specialized processors, eliminating the need for manual intervention while maintaining high accuracy.
Solution Approach 2:
The patent replaces the mechanical manual identification process with an automated AI-based system. Instead of radiologists manually identifying and cropping body regions, AI processors use machine learning models to automatically detect body regions, determine orientations, and route images to appropriate specialized AI processors, substituting human mechanical work with automated computational processes.
2Measurement precision
If multiple AI processors are used to evaluate different body regions, then measurement precision is improved, but device complexity increases due to the need to manage numerous specialized processors
Solution Approach 1:
The system implements universality by creating a modular AI processor architecture where each processor is specialized for a specific body region but follows a unified interface and workflow. The generic workflow engine can route images to any combination of specialized processors based on detected body regions, allowing the system to scale by adding specialized processors without increasing overall system complexity.
Solution Approach 2:
The patent segments the image analysis task into independent specialized AI processors for different body regions (thorax, abdomen, limbs, etc.). Each processor contains trained models specific to its body region, allowing parallel processing of different regions while maintaining modularity. This segmentation enables precise condition detection for each region while keeping individual processor complexity manageable.
3Manufacturing precision
If old AI models are replaced with new models during retraining, then manufacturing precision is improved, but loss of information occurs because valuable data from old models is discarded
Solution Approach 1:
The system merges old and new AI models by implementing an ensemble architecture where both models coexist and their predictions are combined. The workflow engine can execute both the legacy model and the newly trained model on the same images, then aggregate their results through voting, averaging, or other combination strategies, preserving valuable diagnostic data from the old model while incorporating improvements from the new model.
Solution Approach 2:
The patent implements discarding and recovering by maintaining archives of old AI models and their training data. Instead of permanently discarding legacy models during retraining, the system recovers their value by continuing to use them in parallel with new models, allowing comparison of results and preservation of diagnostic patterns that the old models captured but the new models may not yet have learned.
Data Source
AI summary
Systems and methods are described for obtaining an additive AI result from a digital file, the method including: processing the digital file by at least one first artificial intelligence (AI) classifier and at least one second AI classifier thereby obtaining a first evaluation result and at least one second evaluation result respectively; directing the first evaluation result and the at least one second evaluation result to at least one synthesis processor; and comparing the first evaluation result and the at least one second evaluation result to at least one dataset cluster thereby obtaining the additive AI result.