Related Experiment Video
Updated: Aug 1, 2025
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
12.9K
Fast-Training Deep Learning Algorithm for Multiplex Quantification of Mammalian Bioproduction Metabolites via
Anjana Hevaganinge1, Callie M Weber1, Anna Filatova1
1Fischell Department of Bioengineering, University of Maryland, 8278 Paint Branch Dr, College Park, Maryland 20742, United States.
ACS Omega
|May 1, 2023
Summary
A new contactless sensor uses short-wave infrared hyperspectral imaging and deep learning to monitor glucose and lactate in bioreactors. This technology offers precise, real-time metabolite detection without probe fouling, improving bioprocess control.
Area of Science:
- Biotechnology
- Spectroscopy
- Machine Learning
Background:
- Biopharmaceutical manufacturing requires real-time, contactless metabolite monitoring for bioreactor control.
- Existing spectral sensors face limitations like probe fouling and submerged operation.
- Short-wave infrared (SWIR) hyperspectral (HS) imaging offers a potential contactless solution for spectral data collection.
Purpose of the Study:
- To develop an interpretable deep learning system for contactless, real-time metabolite quantification in bioreactor spent media.
- To assess the performance of the developed system for detecting glucose and lactate concentrations.
Main Methods:
- Development of a Convolution Metabolite Regression (CMR) deep learning model.
- Utilizing label-free, contactless SWIR HS images of Chinese hamster ovary (CHO) cell-free spent media.
- Training the CMR system using a dataset of under 500 HS images.
Main Results:
- The CMR system achieved RMSE values of 27 mg/dL for glucose and 20 mg/dL for lactate.
- Performance is comparable to conventional Raman spectroscopy probes (26 mg/dL for glucose, 18 mg/dL for lactate).
- The CMR system trains efficiently within 10 epochs, employing an interpretable architecture.
Conclusions:
- The developed contactless deep learning sensing system enables accurate, real-time metabolite detection.
- This technology can enhance the safety and efficiency of bioreactor process control.
- The interpretable CMR approach facilitates reliable metabolite sensing and spurious prediction removal.
