Glucose Device Batch Analysis Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing manufacturing processes for glucose and other health devices are time-consuming and lack clarity in testing outcomes, limiting their efficiency and accuracy in determining device releasability.
Innovation Solution
A system utilizing machine learning techniques to analyze a subset of devices, leveraging manufacturing and testing data to determine performance information and characterize device batches as releasable, with the option for further testing based on initial results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional testing methods are used to determine device releasability, then manufacturing accuracy and reliability are maintained, but manufacturing time and process complexity increase significantly
Solution Approach 1:
The system performs preliminary analysis of manufacturing data during the manufacturing process itself, rather than waiting for final testing. By evaluating manufacturing parameters and predicting device performance in advance, the system determines releasability before completion of traditional testing, thereby reducing manufacturing time while maintaining reliability
Solution Approach 2:
The patent replaces the mechanical/physical testing system with an information processing system that uses machine learning models to predict device performance. Instead of physically testing each device, the system substitutes computational analysis of manufacturing data to determine releasability, significantly reducing time consumption
2Reliability
If comprehensive testing is performed on all devices, then device performance reliability is ensured, but manufacturing complexity and resource consumption increase
Solution Approach 1:
The system extracts only the critical manufacturing parameters and data points that are most indicative of device performance, rather than analyzing all possible manufacturing data. By selecting and focusing on key features, the system maintains reliability while reducing manufacturing process complexity
Solution Approach 2:
The patent transforms manufacturing data into predictive performance parameters through machine learning models. By changing the parameter representation from raw manufacturing measurements to predicted performance metrics, the system simplifies the evaluation process while ensuring device reliability
3Reliability
If traditional testing methods are used, then device performance is verified, but the outcome clarity and decision-making efficiency decrease
Solution Approach 1:
The system implements a feedback mechanism where manufacturing data is continuously analyzed and compared against learned patterns from training data. The machine learning model provides clear predictive outcomes with confidence levels, enabling decisive go/no-go decisions for device releasability, thereby improving outcome clarity while maintaining verification reliability
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Systems and methods for glucose device batch analyses. An example method includes obtaining analysis datasets reflecting manufacturing information associated with medical devices included in a device batch, the manufacturing data measuring information associated with individual manufacturing steps, and with the analysis datasets being formatted for input into a machine learning model, and with unique identifying information associated with the medical devices being used to aggregate analysis datasets specific to the medical devices. The analysis datasets are provided to the machine learning model, with the machine learning model being trained to output performance information for individual medical devices indicating whether they are in releasable condition. The machine learning model was trained based on ground truth associated with manufactured devices, and the ground truth was derived based on test information associated with the manufactured devices. An interactive user interface is generated which presents summary information associated with the performance information.