Related Experiment Video
Updated: Jun 7, 2026
07:59
Bioluminescence Imaging to Detect Late Stage Infection of African Trypanosomiasis
Published on: May 18, 2016
7.9K
Classification of Trypanosoma brucei mammalian life cycle stages using Deep Learning Algorithms.
Hamid Cheraghi1,2, Lara López-Escobar3, José Rino3
1Department of Biological Physics, Eötvös Loránd University (ELTE), Budapest, Hungary.
Plos Neglected Tropical Diseases
|August 14, 2025
Summary
We developed a deep learning pipeline to classify Trypanosoma brucei parasite forms from unlabeled images. This automated method accurately distinguishes slender and stumpy forms, aiding parasite biology research.
Area of Science:
- Parasitology
- Computational Biology
- Biomedical Imaging
Background:
- Accurate classification of Trypanosoma brucei bloodstream forms (slender and stumpy) is crucial for understanding parasite biology and disease transmission.
- Current methods often rely on fluorescently tagged parasites, limiting scalability and requiring specialized techniques.
- Visual distinction between T. brucei slender and stumpy forms is challenging, necessitating objective classification approaches.
Purpose of the Study:
- To develop and validate a semi-automated deep learning pipeline for segmenting and classifying T. brucei bloodstream forms from unlabeled microscopic images.
- To assess the performance of the developed pipeline against traditional classification methods and other deep learning architectures.
- To establish a scalable and efficient framework for high-throughput analysis of single-cell morphology.
Main Methods:
- A two-stage deep learning pipeline was implemented, starting with Cellpose for parasite segmentation and artifact removal.
- A classification model based on the Xception architecture was employed for distinguishing between slender and stumpy T. brucei forms.
- The classification model was optimized using transfer learning and fine-tuning on unlabeled microscopic images.
Main Results:
- The deep learning pipeline successfully segmented and classified T. brucei bloodstream forms with 97% accuracy.
- The Xception-based model outperformed other standard deep learning architectures, including InceptionV3, ResNet50, and VGG16.
- The developed framework demonstrated effectiveness in analyzing unlabeled morphological data for parasite classification.
Conclusions:
- Deep learning offers a powerful and efficient approach for automated parasite stage classification, overcoming limitations of traditional methods.
- The developed semi-automated pipeline provides a scalable solution for high-throughput analysis in Trypanosoma research.
- This framework is adaptable for various single-cell classification tasks utilizing unlabeled morphological data, advancing automated cell analysis.
