Medical Data Processing Apparatus Standardizing Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning based on medical data from various diagnostic apparatuses faces fluctuations due to differences in installation facilities and apparatus versions, leading to statistical bias and decreased accuracy.
Innovation Solution
A medical data processing apparatus that outputs medical diagnostic data alongside transform-standardized medical data, where the latter is standardized for machine learning without performing part or all of the predetermined processing, thereby reducing statistical bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If medical data from various diagnostic apparatuses is used for machine learning, then the quantity of training data increases, but statistical bias occurs due to fluctuations from differences in installation facilities and apparatus versions
Solution Approach 1:
A standardized data format acts as an intermediary between diverse medical diagnostic apparatuses and the machine learning system. The conversion unit transforms medical data from various apparatuses into a unified standardized format, eliminating statistical bias while preserving the quantity and diversity of training data. This mediator layer allows the system to ingest data from multiple sources without directly exposing the machine learning model to apparatus-specific variations.
Solution Approach 2:
The system changes the parameters of medical data by converting it into a standardized format that removes apparatus-specific variations. The conversion unit modifies data parameters such as imaging conditions, protocol settings, and facility-specific parameters into a unified representation, thereby eliminating statistical bias while maintaining the essential diagnostic information needed for machine learning training.
2Reliability
If standardized medical data is generated without performing predetermined processing, then statistical bias is reduced for machine learning, but the workflow for medical diagnosis must be maintained separately
Solution Approach 1:
The data processing workflow is segmented into two distinct pathways: one for generating standardized medical data optimized for machine learning, and another for generating medical diagnostic data with full predetermined processing for clinical use. The conversion unit handles the standardized data stream while the diagnostic data stream undergoes complete processing, allowing both workflows to operate independently without interfering with each other.
Solution Approach 2:
The system performs preliminary conversion of medical data into standardized format at the point of acquisition, before the data is used for machine learning training. This preliminary standardization action eliminates statistical bias upfront, allowing the machine learning system to receive pre-processed, bias-free data without requiring complex post-processing or data cleaning steps later in the workflow.
Data Source
AI summary
In one embodiment, a medical data processing apparatus includes processing circuitry. The processing circuitry obtains medical data relating to a subject, and outputs medical diagnostic image data obtained by performing predetermined processing on the medical data, along with standardized medical image data based on the medical data, the standardized medical image data being standardized for machine learning without performing part or all of the predetermined processing.