ITEM

Inverse Transformed Encoding Models

This repository contains SPM-compatible MATLAB code for estimating inverse transformed encoding models (ITEM) based on first-level general linear models (GLMs) for functional magnetic resonance imaging (fMRI) data [1,2]. The ITEM approach allows trial-wise linear decoding of discrete experimental conditions (classification) or continuous modulator variables (reconstruction) from multivariate fMRI signals.

An ITEM analysis would usually proceed in two steps:

1. Construct trial-wise design matrix AND estimate trial-wise response amplitudes using ITEM_est_1st_lvl.
   
2. a) Classify discrete experimental conditions from trial-wise parameter estimates via ITEM_dec_class(_SL) OR
   b) Reconstruct continuous modulator variable from trial-wise parameter estimates via ITEM_dec_recon(_SL).

The functions ITEM_dec_class and ITEM_dec_recon are written for ROI-based ITEM analysis, whereas the functions ITEM_dec_class_SL and ITEM_dec_recon_SL serve searchlight-based ITEM analysis. Type help ITEM_fct_name for information on input parameters of these functions. You may also use the review function by typing ITEM_review and selecting an SPM.mat in order to check intermediate results at any time. The code in this repository references some functionality from the MACS toolbox [4]. If this toolbox is on the path, functions not starting by ITEM_ are not required.

This software is in beta testing. Future improvements will include a user interface via SPM's batch editor and a software manual. Preliminary documentation can be found in a preprint uploaded to bioRxiv [1] and a paper published in NeuroImage [2]. In case of questions on the methodology or issues with the toolbox, please contact the corresponding author [3].

[1] https://www.biorxiv.org/content/10.1101/610626v3
[2] https://www.sciencedirect.com/science/article/pii/S1053811919310407
[3] mailto:joram.soch@bccn-berlin.de
[4] https://github.com/JoramSoch/MACS