Fig. 4

Generalizability performance assessment of multimodal deep learning against individual networks. We compared the performance of the MDL architecture against each individual network (i.e. temporal, textual, and demographics) for each system. We calculated 1000 bootstrap samples from the test set. For each sample, we calculated AUC. Finally, for each prediction task and system, we performed a t-test to assess significance between the model that contained all three data types and the models using a single data type; we indicate significance with an asterisk. We corrected for multiple hypothesis testing using Bonferroni correction by multiplying each p value by three. AUC, Area Under the Curve; MDL, Multimodal Deep Learning