Three-Dimensional Convolutional Autoencoder Extracts Features of Structural Brain Images With a "Diagnostic Label-Free" Approach: Application to Schizophrenia Datasets.

In comparison to the conventional method, the 3D-CAE model with 3 blocks showed a trend toward a smaller prediction error for positive symptoms than the ROI method ( p = 0.088; Table 2 ), the mean RMSE (SD) was 4.67 (0.09) and 4.72 (0.04), respectively, suggesting that the 3D-CAE might be comparable or better than the ROI method.

also in 53,322 other papers