Smartphone-Based Biochemical Analyzer Using Deep Learning for Test Strip Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing smartphone-based diagnostic systems for colorimetric tests face challenges with ambient light variations, uncalibrated cameras, and the need for peripheral hardware attachments, leading to reliability issues and user unfriendliness.
Innovation Solution
A system that uses a deep learning method based on convolutional neural networks to learn and adapt to various illumination and positioning conditions, eliminating the need for hardware attachments and providing accurate, standardized measurements by processing images of test strips within a smartphone app.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If photometric devices are used to improve measurement accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent replaces complex photometric devices with a smartphone camera system. The camera captures images of the test strip, and machine learning algorithms process these images to extract quantitative measurements. This substitution eliminates the need for specialized photometric hardware while maintaining or improving measurement accuracy through computational analysis.
Solution Approach 2:
The patent uses digital image copying of the test strip results. Instead of requiring direct optical measurement with specialized devices, the system creates a digital copy (image) of the test strip and analyzes it computationally. This approach simplifies the physical measurement system while enabling accurate quantification through image processing and machine learning.
2Reliability
If peripheral hardware attachments are added to control ambient light, then reliability is improved, but ease of operation deteriorates
Solution Approach 1:
The patent implements self-service through machine learning algorithms that automatically adapt to varying ambient light conditions. The system captures reference images of the test strip under current lighting conditions and uses these to normalize and correct the measurement. This eliminates the need for users to manually control or stabilize lighting environments, making the system both reliable and easy to use.
Solution Approach 2:
The patent dynamically adjusts measurement parameters based on captured images. The machine learning model processes the actual lighting conditions and adapts the analysis parameters accordingly, rather than requiring fixed, controlled lighting conditions. This allows the system to maintain reliability across varying environmental conditions without requiring user intervention.
3Adaptability or versatility
If machine learning algorithms are used to adapt to illumination variations, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary action by capturing reference images of the test strip under the actual measurement conditions before final analysis. These reference images are processed to establish baseline characteristics and normalization factors that are then applied to the quantitative measurement. This preliminary adaptation to lighting conditions simplifies the subsequent analysis and enables the system to handle varying illumination without complex real-time adjustments.
4Measurement precision
If standardized quantitative measurements are achieved through software processing, then measurement precision is improved, but loss of information increases
Solution Approach 1:
The patent implements feedback through an iterative machine learning process. The system captures images, processes them through the trained model, and uses the results to refine and update the model continuously. This feedback loop ensures that the system learns from actual measurement data and environmental variations, improving precision while preserving relevant information through adaptive model updating rather than losing information in static processing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves ergonomic, user-friendly operation with minimized failure rates and increased reliability, providing consistent and accurate quantitative measurements for metabolite quantification in biological samples, such as urine, without the need for external hardware.
Implementation Method 1
They contain test and calibration regions where the latter region is used as the reference to the selected parameter of the tested body fluid (analyte). Dipsticks can be sensitive to a chemical, or enzymatic colorimetric reactions.
Implementation Method 2
Each pad on a dipstick uses an absorbent material that has been impregnated with a mixture of dyes that shift color when contacted to the solution being tested.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The present invention relates to a biochemical analyzing method for quantifying metabolites in a biological sample. The method is based on generating and learning the appearance of a test strip (dipstick) under various conditions to estimate the value/label for an unknown sample image. The method consists of two parts: a training part and a testing part. In the training part, metabolite quantities of a test strip are measured by a biochemistry analyzer; a set of images of the same test strip are captured by a device simulating various ambient lighting conditions; a machine learning model is trained using the images of the test strip and its corresponding metabolite quantities and the learning model is transferred to a smart device. In the testing part, images of a test strip to be analyzed are captured by the smart device and the images are processed using the learning model determined in the training part.