Figure 3
From: Identification of tophi in ultrasound imaging based on transfer learning and clinical practice
Schematic diagram showing the workflow of training and testing. After cropping the region of interest in 375 ultrasound images, the images were assigned to training or test datasets. Data augmentation based on histogram equalization and horizontal flipping was used to increase the number of images in the training dataset. Transfer learning was performed using three pre-trained models with fivefold cross-validation prior to the analysis of diagnostic performance using the test set.
