Related Experiment Video
Updated: Jun 6, 2026
07:40
Microstructured Devices for Optimized Microinjection and Imaging of Zebrafish Larvae
Published on: December 8, 2017
Dual-modal deep learning model with microstrip isoelectric focusing and imaging strategy for feed classification
Youli Tian1, Xiunan Lu2, Yiren Cao1
1School of Automation and Intelligence Sensing, Shanghai Jiao Tong University, Shanghai, 200240, China.
Analytica Chimica Acta
|June 4, 2026
Summary
A new dual-modal deep learning framework accurately identifies animal-derived feed ingredients using microstrip isoelectric focusing (mIEF) and digital imaging. This cost-effective method significantly improves feed safety and traceability.
Area of Science:
- Analytical Chemistry
- Biotechnology
- Machine Learning
Background:
- Accurate identification of animal-derived feed ingredients is crucial for livestock safety and disease prevention.
- Conventional methods like PCR and microscopy have limitations including high cost, low throughput, and lack of universality.
- Single-modal approaches have inherent weaknesses, necessitating a strategy that combines multiple data types for robust classification.
Purpose of the Study:
- To develop an intelligent analytical strategy that synergizes multidimensional information for robust classification of animal-derived feed ingredients.
- To create a dual-modal deep learning framework integrated with an array-format microstrip isoelectric focusing (mIEF) device.
- To overcome the limitations of single-modal analytical approaches in feed ingredient identification.
Main Methods:
- Established a dual-modal deep learning framework using an array-format microstrip isoelectric focusing (mIEF) device.
- Simultaneously captured microscopic biochemical fingerprints (mIEF profiling) and macroscopic physical textures (digital RGB imaging).
- Constructed a dual-branch ResNet18 model with a channel attention mechanism for adaptive fusion of the two modalities.
Main Results:
- Achieved a superior classification accuracy of 96.53% across 15 feed categories.
- Significantly outperformed single-modal baselines: mIEF-only (83.47%) and image-only (49.33%).
- The integrated device supports parallel processing of 12-24 samples per run at an estimated cost of less than $1.5 per sample.
Conclusions:
- Presented a practical analytical strategy for feed traceability by integrating mIEF-based protein fingerprinting with digital image analysis.
- The combined biochemical and visual information improved classification performance compared to single modalities.
- The proposed platform offers a feasible, rapid, accessible, and cost-effective approach for routine screening of animal-derived feed ingredients.
