Related Experiment Video
Updated: Jan 13, 2026
09:34
A Virtual Machine Platform for Non-Computer Professionals for Using Deep Learning to Classify Biological Sequences of Metagenomic Data
Published on: September 25, 2021
4.4K
Advancing biomedical waste classification through a hybrid ensemble of deep Learning, reinforcement Learning, and
Surajet Khonjun1, Rapeepan Pitakaso1, Thanatkij Srichok1
1Artificial Intelligence Optimization SMART Laboratory, Industrial Engineering Department, Faculty of Engineering, Ubon Ratchathani University, Ubon Ratchathani 34190, Thailand.
Waste Management (New York, N.Y.)
|October 28, 2025
Summary
This study introduces the BioSorter, an AI model for classifying pharmaceutical and biomedical waste. It significantly improves accuracy and efficiency in managing infectious waste, reducing costs and enhancing healthcare sustainability.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence in Healthcare
- Waste Management Technology
Background:
- Pharmaceutical and biomedical waste management presents significant challenges, especially for infectious materials.
- Safe and cost-effective disposal methods are crucial for healthcare sustainability.
- Current waste classification methods may lack the necessary accuracy and efficiency.
Purpose of the Study:
- To develop a novel classification model for pharmaceutical and biomedical waste.
- To create an automated system, the "Biosorter," for differentiating infectious from non-infectious waste.
- To enhance the accuracy, efficiency, and usability of medical waste management.
Main Methods:
- A double heterogeneous ensemble model integrating deep learning, reinforcement learning, and differential evolution algorithms.
- Utilized image augmentation, ensemble image segmentation (U-Net, Mask R-CNN, DeepLab V3+), and ensemble convolutional neural network (CNN) architectures (Inception V3, ResNet50, MobileNetV2, DenseNet121).
- Incorporated decision fusion techniques with reinforcement learning and differential evolution for classification.
Main Results:
- The BioSorter model demonstrated superior performance compared to existing deep learning architectures on proprietary and benchmark datasets, achieving accuracy improvements of 5.35% and 9.05%.
- Real-world deployment showed 98% sorting accuracy and a 50% increase in processing throughput.
- Achieved a high System Usability Scale (SUS) score of 93.5, indicating excellent user-perceived efficiency and simplicity.
Conclusions:
- The developed BioSorter represents a significant advancement in pharmaceutical and biomedical waste management.
- The model offers potential for substantial reductions in disposal costs within healthcare environments.
- This technology can enhance overall sustainability and safety in healthcare waste handling.
