Figure - PMC

Skip to main content

An official website of the United States government

Here's how you know

Here's how you know

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

View full-text article in PMC

. 2021 Nov 8;24(12):1772–1779. doi: 10.1038/s41593-021-00947-w

Search in PMC
Search in PubMed
View in NLM Catalog
Add to search

© The Author(s) 2021, corrected publication 2023

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

PMC Copyright notice

Fig. 1 — a, Manually delineated eye masks superimposed on a T1-weighted structural template (Colin27) at MNI coordinate Z = –36. b, Eyeball MR signal reflects gaze direction. The normalized MR signal of eye mask voxels of a sample participant who fixated on a target on the left (X, Y = –10, 0°), right (10, 0°), top (0, 5.5°) or bottom (0, –5.5°) of the screen are plotted. Source data are provided. c, CNN architecture. The model takes the eye mask voxels as three-dimensional (3D) input and predicts gaze position as a two-dimensional (2D; X, Y) regression target. It performs a series of 3D convolutions (3D Conv) with group normalizations (GroupNorm) and spatial downsampling via average pooling (AvgPool) in between. Residual blocks (ResBlock) comprise an additional skip connection. The model is trained across participants using a combination of two loss functions: (1) the Euclidean Error (EE) between the predicted and the true gaze position and (2) the error between the EE and a predicted error (PE). It outputs gaze position and the PE as a decoding confidence measure for each repetition time (TR). d, Schematics of viewing priors. We trained and tested the model on data from 268 participants performing fixations¹¹, smooth pursuit on circular or star-shaped trajectories^12–14 and free viewing¹⁵.

Source data