Deep learning for computational cytology: A survey
Hao Jiang1, Yanning Zhou2, Yi Lin1
1Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China.
Medical Image Analysis
|December 1, 2022
Summary
Deep learning (DL) advances computational cytology for cancer screening by analyzing digitized images. This survey covers DL methods, datasets, applications like cell classification and segmentation, and future research directions.
Area of Science:
- Medical image computing
- Computational pathology
- Artificial intelligence in oncology
Background:
- Computational cytology analyzes digitized cytology images for cancer screening.
- Deep learning (DL) has shown significant achievements in medical image analysis.
- There is a growing body of research on DL applications in cytological studies.
Purpose of the Study:
- To survey over 120 publications on DL-based cytology image analysis.
- To investigate advanced DL methods and comprehensive applications in computational cytology.
- To discuss current challenges and future research directions in the field.
Main Methods:
- Introduction to various deep learning schemes: fully supervised, weakly supervised, unsupervised, and transfer learning.
- Systematic summarization of public datasets and evaluation metrics used in DL for cytology.
- Categorization of versatile cytology image analysis applications.
Main Results:
- DL methods are increasingly applied to diverse cytology image analysis tasks.
- Key applications include cell classification, slide-level cancer screening, and nuclei/cell detection and segmentation.
- A comprehensive overview of existing DL approaches, datasets, and evaluation strategies is presented.
Conclusions:
- Deep learning is a powerful tool revolutionizing computational cytology and cancer screening.
- Further research is needed to address current challenges and explore new potential directions.
- The field requires continued investigation into advanced DL techniques and their clinical translation.
