Fig 1.
The proposed framework for gastric gland segmentation, classification and gastric WSI analysis.
Fig 2.
The proposed GAGL-VTNet model for gastric gland segmentation and classification.
Fig 3.
Deformable transformer encoder consisting of a Multi-scale Deformable Attention Module (MSDAM) and a Feed-Forward Network (FFN).
Table 1.
Performance comparison to other methods.
Fig 4.
Gland segmentation results of GAGL-VTNet model on the GlaS challenge dataset in comparison with ground truth: Yellow color (true positive), red color (false positive), green color (false negative).
Fig 5.
Gland segmentation results of GAGL-VTNet model on GAGL dataset in comparison with ground truth: Yellow color (true positive), red color (false positive), green color (false negative).
Fig 6.
Qualitative results of the segmentation of the gastric tissue, gastric mucosa and gastric glands: a) input 10x H&E-stained image, b) gastric tissue segmentation, c) gastric mucosa identification, d) gastric glands detection.
Table 2.
Comparison of gland classification using different models.
Fig 7.
Gland classification using the GAGL-VTNet model on four WSI: a-b) normal cases, c-d) IM cases. Blue color denotes the glands that have been detected as normal and red color denotes the glands that have been detected as IM glands.
Fig 8.
Box plots comparing the average area that glands cover per WSI between normal, Gastric Atrophy (GA) and Intestinal Metaplasia (IM) cases included in GAGL dataset.
For the comparison the average area of normal glands is used as reference. For IM cases, average area of glands per IM case and average area of glands classified as IM is estimated.
Fig 9.
Box plots comparing the ratio of number of glands to gastric mucosa between normal, Gastric Atrophy (GA) and Intestinal Metaplasia (IM) cases included in the GAGL dataset.
Fig 10.
Box plots comparing the ratio of the area covered by glands to gastric mucosa between normal, Gastric Atrophy (GA) and Intestinal Metaplasia (IM) cases included in the GAGL dataset.