Blood Pressure Estimation via Multi-Path Feature Differences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wearable devices face challenges in accurately monitoring blood pressure due to noisy calibration windows and imbalanced training data, leading to inaccurate predictions, especially when dealing with infrequent extreme changes in blood pressure.
Innovation Solution
A method and apparatus for estimating blood pressure using a wearable device that combines baseline measurements with intermediate and final feature differences, employing a machine learning model to predict blood pressure changes by analyzing feature differences across multiple time windows, thereby reducing the impact of noisy data and imbalanced training sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional single-path blood pressure estimation is used, then the measurement process is simple, but prediction accuracy deteriorates due to noisy calibration windows and imbalanced training data
Solution Approach 1:
The patent segments the blood pressure estimation process into multiple independent paths, each calculating blood pressure changes from different time windows (first path: initial to final time window; second paths: intermediate time windows to final time window). This segmentation allows the system to process diverse data samples through different routes, reducing the impact of noisy calibration windows and imbalanced training data on overall prediction accuracy.
Solution Approach 2:
The patent merges the results from multiple estimation paths by combining the first blood pressure change estimate with multiple second blood pressure change estimates. This merging process integrates information from various time windows and data samples, producing a more robust and accurate final blood pressure prediction that compensates for weaknesses in individual paths.
2Measurement precision
If multiple paths of feature differences are combined, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The computational workload is segmented into multiple independent path calculations that can be performed separately. Each path processes feature differences from specific time windows independently, allowing for optimized resource allocation and potential parallel processing, thereby managing computational complexity while maintaining improved prediction accuracy.
Solution Approach 2:
The system performs partial calculations for each path, computing only the necessary feature differences and blood pressure changes required for that specific path. This avoids redundant computations and excessive processing, balancing the need for multiple paths with the constraint of computational power availability.
Data Source
AI summary
A method includes obtaining a baseline blood pressure at an initial time window; estimating a plurality of intermediate blood pressure change estimates, the intermediate blood pressure change estimates correspond to respective time windows that are subsequent to the initial time window; estimating a final blood pressure change estimate between the initial and final time windows; and obtaining the blood pressure by adding the baseline blood pressure to the final blood pressure change estimate. Estimating the final blood pressure includes estimating a first blood pressure change between the initial time window and the final time window; estimating a plurality of second blood pressure changes, each second blood pressure change is between a respective time window of the respective time windows and the final time window; and estimating the final blood pressure change estimate as a combination of the first blood pressure change and the plurality of second blood pressure changes.