Armen Stepanyants

This paper introduces a user-assisted nonrigid registration method that improves accuracy by incorporating fiducial points, achieving sufficient biomedical application accuracy within minutes.

The paper introduces a normalized level set model to improve segmentation of low-contrast objects in 2D and 3D biomedical images.

This paper shows that noise in neural networks enhances memory recall and suggests that noisy neurons should have higher synaptic weights for optimal memory storage.

This study shows that associative learning in recurrent neural networks can replicate key structural and dynamical properties of local cortical circuits.

The paper develops and evaluates neural network filters to improve 3D optical microscopy images of neurites, showing significant improvements in image quality metrics compared to conventional filters.

This paper introduces a feature-based registration method for 3D time-lapse microscopy images, achieving sub-voxel accuracy in tracking axonal structures over time.