Accurate Diagnosis and Survival Prediction of Bladder Cancer Using Deep Learning on Histological Slides
Problem
Early diagnosis and survival prediction of bladder cancer from whole-slide digitised histological images, to reduce mortality.
Why it matters
Early bladder cancer diagnosis and treatment are essential to reduce mortality.
What the AI did
Two weakly supervised deep learning models were trained on 926 WSIs from TCGA: BlcaMIL for diagnosis and MibcMLP for survival prediction in muscle-invasive bladder cancer.
Results
BlcaMIL identified bladder cancer with accuracy 0.987 in external validation (250 WSIs, 150 patients), comparable to expert uropathologists and outperforming a junior pathologist.
Impact
Externally validated on an independent hospital cohort, supporting AI-assisted diagnosis.
Record
| Venue | Cancers |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Qingyuan Zheng, Rui Yang, Xinmiao Ni, Song Soo Yang, Lin Xiong, Dandan Yan, Lingli Xia, Jingping Yuan et al. |
| Citations | 51 (OpenAlex, at import time) |
| References | 74 |
| Model or method | BlcaMIL, MibcMLP |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/cancers14235807 |
| Open access | gold |
| Keywords | bladder cancer, deep learning, whole slide images, diagnosis, survival prediction |
Importance rating
significant — thousands to millions affected
Early bladder cancer diagnosis and treatment are stated as essential to reduce mortality; externally validated diagnostic model comparable to expert uropathologists, but no disease-scale numbers given.