Robust whole slide image analysis for cervical cancer screening using deep learning
Problem
Current algorithms perform poorly on whole slide image analysis for cervical cancer screening, failing to generalize across staining and imaging conditions.
Why it matters
Cervical cancer screening needs to scale up, and computer-assisted diagnosis is key to reaching more patients.
What the AI did
A progressive lesion cell recognition method combining low- and high-resolution WSIs plus a recurrent neural network-based classifier evaluates lesion degree of whole slide images.
Results
93.5% specificity and 95.1% sensitivity on 1,170 multi-center independent test WSIs, comparing favourably to three independent cytopathologists; 88.5% true positive rate for top 10 lesion cells on 447 positive slides
Impact
Validated on multi-center independent test sets with performance comparable to cytopathologists; deployment mentioned but abstract truncated.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Shenghua Cheng, Sibo Liu, Jingya Yu, Gong Rao, Yuwei Xiao, Wei Han, Wenjie Zhu, Xiaohua Lv et al. |
| Citations | 195 (OpenAlex, at import time) |
| References | 55 |
| Model or method | Recurrent neural network-based WSI classification |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-021-25296-x |
| Open access | gold |
Importance rating
significant — thousands to millions affected
Abstract states computer-assisted diagnosis is key for scaling up cervical cancer screening; multi-center validation, cytopathologist-level performance, and deployment described.