Deep Learning Methods for Lung Cancer Segmentation in Whole-Slide Histopathology Images—The ACDC@LungHP Challenge 2019
Problem
Automatically segmenting lung cancer tissue pixel-wise in whole-slide histopathology images to support diagnosis.
Why it matters
Accurate segmentation of lung cancer in pathology slides is a critical step in improving patient care.
What the AI did
Ten deep-learning challenge methods performed automatic cancer segmentation on an annotated dataset of 200 patient images, evaluated by precision, accuracy, sensitivity, specificity and DICE coefficient.
Results
DICE coefficient ranged 0.7354±0.1149 to 0.8372±0.0858; best method close to inter-observer agreement (0.8398±0.0890).
Impact
Best method neared human inter-observer agreement on the challenge benchmark.
Record
| Venue | IEEE Journal of Biomedical and Health Informatics |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Zhang Li, Jiehua Zhang, Tao Tan, Xichao Teng, Xiaoliang Sun, Hong Zhao, Lihong Liu, Yang Xiao et al. |
| Citations | 157 (OpenAlex, at import time) |
| References | 68 |
| Model or method | deep learning; specific architectures not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1109/jbhi.2020.3039741 |
| Open access | hybrid |
| Keywords | lung cancer, whole-slide imaging, segmentation, deep learning, challenge, DICE coefficient |
Importance rating
significant — thousands to millions affected
Abstract calls accurate lung cancer segmentation a critical step in improving patient care; significance inferred from cancer diagnosis context, no scale stated.