Table 2 Results for the baseline models trained and validated using only wound images from the new multimodal dataset, and tested on the DFUC 2022 test set, which also comprises only wound images. All wound images are \(640 \times 480\) pixels. Ep - epoch; Tr - train; V - validation; Te- test; UN - U-Net; IoU - intersection over union; DSC - Dice similarity coefficient; FPE - false positive error; FNE - false negative error. Note that no pretraining and no pre- or post-processing was used in these experiments.

From: Gaussian random fields as an abstract representation of patient metadata for multimodal medical image segmentation