Bubble Detection in Well Plate Images Using Convolutional Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current light scattering measurements in multiwell plates are affected by background noise from bubbles, contaminants, and other imperfections, leading to inaccurate sample property assessments due to the inability to effectively detect and remove bubbles from images of samples in well plates.
Innovation Solution
A computer-implemented method using a trained convolutional neural network (CNN) processes images of well plates to detect bubbles by receiving images, cropping the well of interest, and applying stochastic gradient descent training to determine the probability of bubble presence, thereby improving the accuracy of light scattering measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If light scattering measurements are performed in multiwell plates, then high-throughput sample analysis is achieved, but measurement precision deteriorates due to background noise from bubbles and contaminants
Solution Approach 1:
The system performs preliminary bubble detection and identification using image processing and machine learning algorithms before the actual light scattering measurements. By detecting bubbles in advance and flagging affected wells, the system prevents contaminated data from being collected, thus maintaining measurement precision while preserving high-throughput capability
Solution Approach 2:
An intermediary image processing system is introduced between the multiwell plate and the light scattering measurement system. This intermediary layer captures images, processes them through CNN algorithms, and provides quality control feedback, thereby filtering out bubbles and contaminants before they affect the measurement precision
2Device complexity
If traditional image processing methods are used to detect bubbles, then device complexity is reduced, but detection precision deteriorates due to inability to effectively distinguish bubbles from sample features
Solution Approach 1:
Traditional mechanical or rule-based image processing methods are replaced with a machine learning-based convolutional neural network system. The CNN automatically learns to distinguish bubbles from sample features by training on labeled images, achieving high detection precision without requiring complex manual feature engineering or adjustment of multiple processing parameters
3Reliability
If manual bubble removal methods are used, then measurement reliability is improved, but productivity deteriorates due to time-consuming manual intervention
Solution Approach 1:
The system performs automated bubble detection, identification, and flagging using machine learning algorithms without requiring manual intervention. The CNN automatically processes images, identifies bubbles, and marks affected samples for exclusion, enabling the system to self-correct measurement data while maintaining high throughput and reliability
Data Source
AI summary
The present disclosure describes an apparatus, method, system, and computer program product of detecting bubbles in images of a sample in wells of a well plate. In an exemplary embodiment, In an exemplary embodiment, the computer implemented method, the system, and the computer program product include receiving, by a computer system, at least one image of at least one well of a well plate containing a liquid sample, executing, by the computer system, a set of logical operations cropping the image to include a well of interest, and executing, by the computer system, a set of logical operations processing, by a trained convolutional neural network, the cropped image, resulting in a probability that the image depicts at least one bubble.