Two new papers on AI in Neuroscience released
Two papers that have been presented by INM-1 PhD students at this year’s International Symposium on Biomedical Imaging (ISBI) are now released at IEEE Explore.
Two papers that have been presented by INM-1 PhD students at this year’s International Symposium on Biomedical Imaging (ISBI) are now released at IEEE Explore.
Denoising Diffusion Probabilistic Models for Image Inpainting of Cell Distributions in The Human Brainby Jan-Oliver Kropp et al
introduces a denoising diffusion probabilistic model extended with the so-called RePaint method, to accurately inpaint missing or corrupted cell distribution data in brain images, generating realistic cellular patterns and statistics. This will help to enhance ultra-high resolution brain images for improved 3D reconstruction and visualisation.
Example patches from the evaluation dataset with present artifacts. Left: Image patches with crystal artefact (top) and missing tissue (bottom). Right: Corresponding patches with artifacts removed by our DDPM and the RePaint method. Red annotations show manually provided annotations of the respective artefact. The repainted images show no signs of the artifacts and closely follow the true cytoarchitecture (e.g., the columnar organization).
Analyzing Regional Organization of The Human Hippocampus in 3D-PLI Using Contrastive Learning and Geometric Unfoldingby Alexander Oberstrass et al
presents a novel method for analyzing the regional organization of the human hippocampus in 3D polarized light imaging (3D-PLI) by integrating geometric unfolding techniques with deep texture features, extracted using a contrastive learning approach, to identify clusters that align with known hippocampal subfields.