2022-21 Deep learning-powered in situ bioprocess monitoring with Raman spectroscopy
Project Description
- Integrate a deep learning model with Raman spectral measurements to enable precise identification of cell culture conditions with minimal error
- Create a spectral data collection of metabolites using Surface-enhanced Raman scattering (SERS)
- Perform offline measurements using Raman fiber probe
- Develop EfficientNet and ResNet-50 algorithms with transfer learning to identify and assess metabolites
How the Project may be Transformative and/or Benefit Society
- SERS technology combined with deep learning-based regression models will permit highly sensitive and label-free multiplexed measurements in bioreactor media
- Reduction of measurement time
- Unique “representation learning” enables direct training of biological fingerprints from raw spectra
- DL works well on noisy data obviating high thresholds on SNRs and, hence, reducing measurement times
- Roadmap for designing SERS-active fiber probes that allow for in situ bioprocess monitoring with requiring additional sampling
Project Pitch
Project Deliverables
Private: 2022-21 Poster
0 Downloads
| Project Leaders | Michael Betenbaugh, Ishan Barman |
|---|---|
| Lead Student | Peng Zheng, Andy Nelson |
| University | Johns Hopkins University |
Mentor Meetings
Milestones
Publication: Deep Learning-Powered Colloidal Digital SERS for Precise Monitoring of Cell Culture Media
1 Download