Mobile Machine-Learning Segmentation for Lateral Flow Test Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated lateral flow assay (LFA) evaluation techniques face challenges such as high manufacturing costs due to marker-based systems, vulnerability to tilt and lighting conditions, limited ability to analyze multiple test cassettes, and high computational requirements, making them inefficient and inaccurate.
Innovation Solution
A computer-implemented method using machine-learning models for image segmentation and analysis on mobile devices, employing downscaled images and separate models for test cassette, viewport, and signal detection, enabling efficient and accurate quantitative evaluation of LFA results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If marker-based systems are used for automated LFA evaluation, then the ability to detect test cassettes is improved, but manufacturing costs increase and device complexity increases
Solution Approach 1:
The patent extracts and removes the fiducial marker component from the test cassette design. Instead of requiring printed markers on each device, the system uses natural geometric features of the cassette housing (corners, edges, viewports) as detection references, eliminating the need for additional manufacturing steps to apply markers
Solution Approach 2:
The patent makes the cassette housing structure serve multiple functions: it provides structural containment, visual identification through geometric features, and serves as the fiducial reference system itself. The viewports and housing geometry become multi-functional elements that aid both sample containment and automated detection without requiring separate marker components
2Difficulty of detecting and measuring
If marker-based systems are used for automated LFA evaluation, then the ability to detect test cassettes is improved, but device complexity increases
Solution Approach 1:
The patent removes the fiducial marker layer from the system, simplifying the test cassette to its essential components. The detection algorithm directly uses the housing geometry and viewport structures without requiring marker recognition, reducing system complexity
Solution Approach 2:
The test cassette housing and viewports serve their primary containment function while simultaneously providing the geometric features needed for automated detection. The structure itself performs the dual role of sample holder and fiducial reference, eliminating the need for separate marker components and their associated manufacturing complexity
3Difficulty of detecting and measuring
If visual markers are used as origin for image recognition, then the ability to locate viewports is improved, but vulnerability to tilt and angle increases
Solution Approach 1:
The patent utilizes the asymmetric geometric arrangement of viewports and housing features to establish orientation and location. The unique spatial relationships between viewports and housing corners provide tilt-invariant geometric constraints that allow the system to determine viewport positions and cassette orientation simultaneously, making the system robust to arbitrary viewing angles
Solution Approach 2:
The patent transitions from using 2D visual markers on the cassette surface to utilizing the 3D geometric structure of the housing and viewport arrangement. By detecting corner points and spatial relationships in three-dimensional space, the system can compute viewport locations and correct for tilt effects through geometric transformation, achieving angle-independent detection
4Measurement precision
If automated evaluation techniques are adapted for each individual test cassette type, then the precision for specific assays is improved, but device complexity and ease of operation worsen
Solution Approach 1:
The patent implements a dynamic, adaptive detection system that automatically adjusts to different test cassette types during runtime. The machine learning model learns from the geometric features presented in each image and adapts its detection parameters accordingly, allowing a single system to handle multiple cassette layouts without pre-programming or manual configuration for each type
Solution Approach 2:
The system dynamically adjusts detection parameters such as search regions, geometric constraints, and analysis thresholds based on the observed cassette geometry and viewport arrangement in each image. This parameter adaptation allows the same core algorithm to achieve high precision across different test cassette types by automatically tuning to the specific geometric characteristics of each assay
5Measurement precision
If extensive processing power is used for image recognition, then the accuracy of LFA evaluation is improved, but productivity and ease of operation worsen due to high computational requirements
Solution Approach 1:
The patent extracts and removes unnecessary computational steps from the image processing pipeline. By using geometric feature detection based on corner points and housing structure rather than full-image analysis, the system achieves high accuracy with significantly reduced computational load, enabling fast processing on mobile devices
Solution Approach 2:
Instead of analyzing the entire image to find test results, the patent inverts the approach by first detecting the cassette geometry and viewports, then using those detected regions to guide the analysis. This region-of-interest approach dramatically reduces the amount of image data that requires intensive processing, improving both speed and energy efficiency while maintaining accuracy
Data Source
AI summary
Computer-implemented methods for use in lateral flow test evaluation. One method comprises obtaining, preferably by a mobile electronic device, a digital image that depicts at least one test cassette, wherein the test cassette comprises at least one viewport and wherein the viewport comprises at least one test indicator. The method may comprise performing, preferably by the mobile electronic device, an image segmentation step to recognize at least one test indicator depicted in the digital image, and performing an evaluation step for producing at least one evaluation result based, at least in part, on the recognized at least one test indicator. The image segmentation step may comprise generating at least one object marker, in particular at least one mask and/or bounding box, based on a downscaled version of the obtained digital image and applying the at least one object marker to the obtained digital image or to a part thereof.