Assay Reading Using Convolutional Neural Networks for Orientation Invariance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic readers for lateral flow tests face challenges such as reliance on single, stationary images, sensitivity to orientation and image quality, inability to read multiple tests simultaneously, and requirement for a priori knowledge of test device dimensions.
Innovation Solution
A computer-implemented method using a trained convolutional neural network to read test regions of assays by analyzing digital image data, estimating intensity values, and classifying assays regardless of orientation or image quality, allowing for simultaneous reading of multiple tests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single stationary image is used for reading assays, then the device complexity is reduced, but the reading accuracy deteriorates due to sensitivity to orientation and image quality
Solution Approach 1:
The system transitions from processing a single stationary image to processing a sequence of images over time. The video data allows the system to dynamically adjust and select the best frame for reading, accommodating variations in orientation and image quality without requiring complex mechanical adjustment mechanisms.
Solution Approach 2:
The system performs preliminary processing of video data to identify and select optimal frames before final reading. By analyzing multiple frames in advance, the system can pre-select frames with appropriate orientation and quality, improving reading accuracy without adding complex real-time adjustment mechanisms.
2Measurement precision
If a priori knowledge of test device dimensions is required, then the measurement precision may be improved, but the adaptability deteriorates
Solution Approach 1:
The system performs self-calibration by automatically determining test device dimensions and characteristics from the video data itself, without requiring pre-programmed knowledge of specific device dimensions. The machine learning model learns to identify and adapt to different device types autonomously.
Solution Approach 2:
The system dynamically adjusts processing parameters based on the actual test device being analyzed. By extracting dimensional information directly from the video data and adapting processing parameters accordingly, the system maintains measurement precision across different device types without requiring fixed a priori knowledge.
3Productivity
If multiple tests are read simultaneously, then the productivity is improved, but the device complexity increases
Solution Approach 1:
The system uses a universal video capture and processing approach that can handle multiple test devices simultaneously through software-based identification and segmentation. The machine learning model automatically distinguishes between different tests in the field of view, enabling multi-test reading without requiring multiple dedicated sensor systems.
4Ease of operation
If manual reading is performed, then the ease of operation is improved, but the reliability deteriorates due to user error
Solution Approach 1:
The system replaces manual visual inspection with automated image analysis using machine learning algorithms. The video-based optical system captures test results, and the trained model automatically interprets the data, eliminating human factors such as vision limitations, bias, and fatigue while maintaining ease of operation through simple video capture.
Data Source
AI summary
A computer implemented method for reading a test region of an assay, the method comprising: (i) providing digital image data of a first assay; (ii) inputting the digital image data into a trained convolutional neural network configured to output a first probability, based on the input digital image data, that a first region of pixels of the digital image data corresponds to a first test region of the assay; (iii) if the first probability is at or above a first predetermined threshold, accepting the first region of pixels as a first region of interest associated with the first test region; and (iv) estimating an intensity value of a portion of the first test region in the first region of interest.