Veterinary Diagnosis Support Using Type-Specific Learned Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image analysis techniques fail to accurately determine the presence of lesions in animals, such as dogs, due to variations in organ and bone shapes and sizes based on body length or weight, making it difficult for veterinarians to effectively diagnose using medical images.
Innovation Solution
A diagnosis support device and method that acquires medical image data and type information classified by body length or weight, using a learned model to determine the presence or absence of abnormalities in medical images, with separate models generated for different animal types, such as small, medium, and large-sized dogs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single learned model is used for all animals regardless of type, then the device complexity is reduced, but the measurement precision of lesion detection deteriorates due to variations in organ and bone shapes across different animal types
Solution Approach 1:
The patent divides the animal population into multiple type groups based on body length and weight categories. Separate learned models are trained for each type group, allowing the system to segment the problem into manageable parts that can be processed with type-specific models, thereby improving lesion detection accuracy without requiring a single overly complex universal model
Solution Approach 2:
The patent applies local quality by training each learned model with medical images specific to its corresponding animal type. Each model develops specialized knowledge of the organ and bone structures characteristic of its target animal type, ensuring that the detection algorithm is optimized for the local anatomical features of each group rather than using a generic one-size-fits-all approach
2Measurement precision
If separate learned models are trained for each animal type, then the measurement precision of lesion detection is improved, but the device complexity increases due to multiple models and data management requirements
Solution Approach 1:
The patent creates a universal diagnosis support device that can handle multiple animal types through a unified system architecture. The device accepts type information as input and automatically selects or switches between the appropriate learned model, providing multi-functional capability without requiring separate standalone systems for each animal type
Solution Approach 2:
The patent introduces type information as an intermediary element that bridges the gap between the input medical image and the appropriate learned model. This intermediary allows the system to route the analysis through the correct specialized model based on the animal's type classification, simplifying the management of multiple models by providing a clear selection mechanism
3Speed
If type information is not considered in the analysis, then the processing speed is maintained, but the reliability of diagnosis deteriorates because organ and bone shapes vary by animal type
Solution Approach 1:
The patent performs preliminary classification of the animal type before conducting the detailed lesion analysis. By determining the animal's type category in advance based on body length and weight information, the system can pre-select the appropriate learned model, ensuring that the subsequent analysis is performed with the correct specialized model without adding significant processing time
Data Source
AI summary
A diagnosis support device acquires medical image data representing a medical image obtained by imaging an animal as a subject with a medical image capturing device and type information representing a type which is classified by at least one of a body length or weight of the animal and to which the subject belongs, and determines presence or absence of an abnormality in the medical image of the subject based on the acquired medical image data and type information and a learned model learned in advance using a set of a plurality of pieces of the medical image data for learning and the type information.