Related Experiment Video
Updated: Jul 30, 2026
07:46
Microfluidic Co-Culture Models for Dissecting the Immune Response in in vitro Tumor Microenvironments
Published on: April 30, 2021
4.8K
Relationship between a deep learning model and liquid-based cytological processing techniques
Katsuhide Ikeda1, Nanako Sakabe1, Sayumi Maruyama1
1Pathophysiology Sciences, Department of Integrated Health Sciences, Nagoya University Graduate School of Medicine, Nagoya, Japan.
Summary
Deep learning models for cell analysis require consistent liquid-based cytology (LBC) processing techniques for accurate detection and classification. Using the same LBC method for training and analysis ensures high performance, while varied methods reduce accuracy.
Area of Science:
- Cytopathology
- Artificial Intelligence
- Deep Learning
Background:
- Liquid-based cytology (LBC) standardizes specimen preparation but introduces variations in cytomorphology based on processing techniques.
- Deep learning (AI) shows promise for cell detection and classification in cytopathology.
Purpose of the Study:
- To investigate the impact of two LBC processing techniques (ThinPrep and SurePath) on AI-driven cell detection and classification.
- To determine the optimal conditions for deep learning models in cytopathology.
Main Methods:
- Cytological specimens were prepared using ThinPrep and SurePath methods.
- Deep learning models (one-cell and five-cell) were trained and tested using these preparations.
Main Results:
- High accuracy in cell detection and classification was achieved when training and detection preparations used identical LBC techniques.
- Models trained on ThinPrep preparations outperformed those trained on SurePath.
- Accuracy significantly decreased when training and detection preparations used different LBC techniques (P < 0.01).
- A model trained on both techniques showed slightly reduced but still high accuracy.
Conclusions:
- Variations in cytomorphology due to different LBC techniques significantly affect deep learning model performance.
- Consistent LBC processing techniques are crucial for accurate AI-based cell detection and classification.
- Developing globally applicable AI models requires training with diverse LBC preparation techniques.
