Packed Cell Volume Prediction Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual and labor-intensive process of determining packed cell volume in bioreactors is time-consuming and prone to low accuracy due to subjective visual estimation, leading to substantial variance between individuals and over time.
Innovation Solution
A non-linear machine learning model, such as a neural network, is used to infer or predict packed cell volume based on measured cell culture characteristics, reducing the need for manual assessments and providing more accurate estimates, which can generate control data for devices in the harvesting process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual estimation is used to measure packed cell volume, then the measurement process is simple and requires basic equipment, but the accuracy is low and there is substantial variance between individuals and over time
Solution Approach 1:
The patent replaces the manual visual estimation process with an automated image analysis system using computer vision and machine learning algorithms. The system captures images of centrifuged samples and automatically calculates packed cell volume, eliminating subjective human judgment and improving measurement precision while maintaining operational simplicity through software-based automation.
Solution Approach 2:
The system creates a digital copy (image) of the physical centrifuged sample and analyzes this copy using automated algorithms. This allows the measurement to be performed on a digital representation rather than requiring direct visual estimation, enabling more precise and consistent measurements without adding physical complexity to the sampling process.
2Productivity
If manual assessment of packed cell volume is performed, then the process requires minimal equipment, but it is very time consuming (roughly three hours for one person assessing samples from a row of eight bioreactors)
Solution Approach 1:
The system enables self-service automation where the image analysis and packed cell volume calculation are performed automatically without requiring manual intervention for each measurement. The automated processing of images and calculations significantly reduces the time required per sample while maintaining measurement quality, thereby increasing overall productivity.
Solution Approach 2:
By replacing the time-consuming manual visual estimation with automated image analysis and machine learning algorithms, the system dramatically increases processing speed. The automated system can analyze multiple samples in parallel, reducing the total time required for assessing packed cell volume across multiple bioreactors from three hours to a fraction of that time.
3Extent of automation
If automated sampling is implemented for cell measurements, then productivity increases, but packed cell volume measurement cannot be automated due to required interaction with centrifuge and visual interpretation
Solution Approach 1:
The patent extracts the visual interpretation step from the manual process and replaces it with automated image analysis software. By separating the image capture function from the interpretation function and automating the latter, the system enables packed cell volume measurement to be integrated into automated sampling workflows while maintaining ease of operation through software-based analysis.
Solution Approach 2:
The system introduces an intermediary layer (image analysis software with machine learning algorithms) between the physical sample and the measurement result. This intermediary automatically processes the visual information from centrifuged samples, enabling automated sampling systems to perform packed cell volume measurements without requiring direct human interaction with the centrifuge or visual interpretation, thereby increasing both automation extent and ease of operation.
Data Source
AI summary
A method of cell culture assessment (e.g., prior to a drug substance harvesting process) includes obtaining a plurality of parameters associated with a cell culture, and inferring or predicting a value or classification indicative of packed cell volume. Inferring or predicting the packed cell volume includes applying the plurality of parameters as inputs to a non-linear machine learning model. The method also includes generating an output indicative of the inferred or predicted value or classification.