Tab.1

MRI findings of all subjects"

Tab.2

MRI findings in the test(validation)set"

Fig.1

Example of raw image(a)and labeled image(b), where the red part is the articular disc,the green part is the condyle,and the blue part is the glenoid fossa"

Fig.2

Flow diagram of 2D model"

Fig.3

Flow diagram of 3D model"

Fig.4

Convergence curve of nnU-Net model training"

Fig.5

MRI image without ADD"

Fig.6

MRI image with ADDwR"

Fig.7

MRI image with ADDwoR"

Tab.3

The Dice, Recall, Precision and F1 score of 2D/3D model in the close or open mouth images"

Tab.4

Results of One-way ANOVA test:there was no significant difference in the segmentation ability of the four models(2D/3D models in the open and closed images)"

Tab.5

Results of One-way ANOVA test:there were statistical differences in the segmentation effect of the three joint tissues(articular disc,condyle,and glenoid fossa)"

Fig.8

Plot of Dice distribution for each singleton of the test set,with each dot representing an outlier"

Tab.6

The diagnostic accuracy of ADD by young doctors based on the original MRI image and the segmentation results of 2D and 3D models(with the senior doctor’s diagnosis for the correct diagnosis) %"

Fig.9

Dice coefficient discretization case example diagram"