Related Experiment Video
Updated: Sep 16, 2025
07:13
Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy
Published on: February 25, 2021
4.0K
Artificial intelligence based hyperspectral biomass estimator for cyanobacteria cultivation
J M Fernández Montenegro1, I Pérez Couñago1, S Iglesias Moreira1
1Smart Systems and Smart Manufacturing (S3M) Department, AIMEN technology centre, Porriño, Spain.
Bioresource Technology
|July 7, 2025
Summary
Hyperspectral imaging and machine learning accurately predict cyanobacterial biomass. A simplified multispectral approach also shows promise for cost-effective, real-time monitoring of photosynthetic bioprocesses.
Area of Science:
- Biotechnology
- Spectroscopy
- Machine Learning
Background:
- Biomass monitoring is crucial for photosynthetic bioprocesses.
- Hyperspectral imaging (HSI) and machine learning (ML) offer potential for non-invasive biomass estimation.
- A limited proof-of-concept exists for linking HSI data to biomass prediction.
Purpose of the Study:
- To establish a proof-of-concept for HSI and ML in cyanobacterial biomass estimation.
- To develop and validate ML models for biomass prediction using spectral and image data.
- To explore the feasibility of low-cost multispectral systems for biomass monitoring.
Main Methods:
- Acquired 450 HSI datasets and 205 biomass measurements from cyanobacteria cultures.
- Utilized a compact push-broom camera for data acquisition.
- Preprocessed images to extract spectral information and trained three ML algorithms (Fully Connected Neural Network, etc.).
Main Results:
- The Fully Connected Neural Network model achieved high accuracy, with a mean absolute error below 4% (37 mg/L).
- A simplified multispectral model using only three wavelengths demonstrated comparable accuracy to HSI.
- Demonstrated the feasibility of using spectral imaging for biomass estimation.
Conclusions:
- HSI combined with ML provides an effective method for cyanobacterial biomass estimation.
- Multispectral imaging presents a viable, potentially lower-cost alternative for biomass monitoring.
- This research supports the development of real-time, non-invasive tools for photosynthetic bioprocesses.
