Table 3 Comparison of best metrics shown in related work compared to those generated from the proposed deep learning model in this paper. Since these are tile datasets, spatial awareness via pyramid tiles was not used for the proposed model. This variant of the proposed model is referred to as HCNN rather than SA-HCNN.
From: Scalable deep learning artificial intelligence histopathology slide analysis and validation
Comparative results on additional datasets | Method | Comparative best result from paper |
|---|---|---|
MEL | (Aubreville et al., 2020) DL | F-score: 0.707 ± 0.013 |
HCNN | F-score: 0.832 | |
TUPAC16 | (Sohail et al., 2021) DL | F-score: 0.72 ± 0.026 |
HCNN | F-score: 0.986 | |
MHIST | (Wei et al., 2021) DL | AUC: 0.841 ± 0.011 |
HCNN | AUC: 0.927 | |
PCam | (Veeling et al., 2018) DL | AUC: 0.963 |
HCNN | AUC: 0.994 |
