Table 3 Results for the HarDNet-CWS model trained and validated using the new multimodal dataset, and tested on the DFUC 2022 test set (image size = \(640 \times 480\) pixels) using GRFs. i - default value of GRF power spectrum integer component; IoU - intersection over union; DSC - Dice similarity coefficient; FPE - false positive error; FNE - false negative error; DOB - date of birth; HDD - health and disability decile. Note that no pretraining, augmentation, or post-processing was used in these experiments.

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