Fig. 10 | Scientific Reports

Fig. 10

From: An interpretable framework for gastric cancer classification using multi-channel attention mechanisms and transfer learning approach on histopathology images

Fig. 10

Confusion matrices for complete GasHisSDB database from three randomized experiments using the proposed MCAM model. (a)-(c) represent results on validation data, while (d)-(f) correspond to results from randomized experiments on the testing dataset. Each column corresponds to one experiment. The green blocks indicate the counts and percentages of true positive and true negative cases, while the red blocks represent false positive and false negative cases. In the last row, the first block shows sensitivity for normal cases and specificity for abnormal cases, the middle block shows sensitivity for abnormal cases and specificity for normal cases, and the last block represents the overall classification accuracy as a percentage. This visualization highlights the model’s consistent performance across all experiments.

Back to article page