Deep Learning Model Training and Evaluation with 5-Fold Cross-Validation

To train the deep learning model and evaluate the results, 1471 images were divided into a training set and a validation set. For each iteration, approximately one-fifth of the images (303 images) were randomly selected from the 1471 images using a 5-fold cross-validation method (Resch et al., 2021) to form the validation set, while the remaining images (1168 images) formed the training set. During the experiment, the images in the training set were used for training, and the images in the validation set were used to test the segmentation method. This evaluation process was repeated 10 times, alternating between the training set and validation set, and the evaluation metrics were calculated each time. The average and standard deviation of each run were calculated to generate the final results.