AI-based association analysis

The distribution of the 2-D latent space for the synthetic images in the test set of Experiment 1, and the eleven reconstructed images sampling along the brightness-related vector, (a) without (i.e., vector 𝑝⃗) and (b) with correction (vector 𝑝∗⃗) for the confounding of circle radius, together with the predicted brightness 𝑡̂ (Eq. 6 and Eq. 7). Z1-axis: the first dimension of the latent space; Z2-axis: the second dimension. Each data point in the latent space represents an input image, which is denoted by its radius and brightness. After training, eleven frames were reconstructed by sampling eleven points along the vector 𝑝⃗ and 𝑝∗⃗ (Eq. 6, Eq. 7) to visualize the confounding effects.

Association between global cognition and brain images with a 3D MRI dataset. Reconstructed supratentorial modulated grey matter maps using the sampled latent features along the direction of increasing g-factor (a) without correcting for confounders, and (b) with correcting for age, sex, and educational years. The results are generated using the semi-supervised model, averaged over the five folds, and masked out the statistically non-significant region. Color bar shows the direction and magnitude of the changes of GM density.