Skip to main content
Fig. 7 | Journal of Translational Medicine

Fig. 7

From: Predicting long-term progression of Alzheimer’s disease using a multimodal deep learning model incorporating interaction effects

Fig. 7

Ablation studies on basic model architecture. A Performance comparison of the models using classical convolution and separable convolution. The models employing classical 3D convolution (Conv3D-based) and separable 3D convolution (SepConv3D-based) were cross-validated with identical settings. Each bar in the chart represents the mean value across folds for respective metrics. B Performance comparison of models based on different fusion schemes. The models based on triple outer product fusion and stepwise fusion were cross-validated under the same settings. Each bar in the chart represents the mean value across folds for respective metrics

Back to article page