ICASSO analysis running on matlab script keeps crashing all the time

[PenghuiDu]

There might be some issues with the memory usage... My PC also went bluescreen several times when trying to run this in batch.

[PenghuiDu]

It's not dependent on MATLAB version (both R2016b and R2025a has the issue), and also not dependent on the platform (same for windows / linux.)

[PenghuiDu]

This is the preference file I'm using when running ICA. The error persist with ICASSO but the occurance seems pretty random, and could crash MATLAB anytime.

`%% Batch script for testing different GIFT ICA algorithms.
set_up;
met_name = 'Glu_abs';
met_name_ = 'Glu_abs';
sub_list = {'sub-08', 'sub-12', 'sub-13'};
sub_n_components = {20, 20, 20};
algo_types = [1, 2, 6, 13];
ica_names = {...
'Infomax+ICASSO', ...
'FastICA+ICASSO', ...
'JADE-OPAC+ICASSO', ...
'ICA-EBM+ICASSO'...
};
prep_folders = {...
'1-remove_mean_per_timepoint', ...
'2-remove_mean_per_voxel', ...
'3-intensity_normalization', ...
'4-variance_normalization'...
};

algo_idx = 2;
sub_idx = 1;
% Data Pre-processing options
% 1 - Remove mean per time point
% 2 - Remove mean per voxel
% 3 - Intensity normalization
% 4 - Variance normalization
preproc_type = 3;

% Enter the values for the variables required for the ICA analysis.
% Variables are on the left and the values are on the right.
% Characters must be enterd in single quotes
%
% After entering the parameters, use icatb_batch_file_run(inputFile);

%% 'Which ICA Algorithm Do You Want To Use';
% see icatb_icaAlgorithm for details or type icatb_icaAlgorithm at the
% command prompt.
% Note: Use only one subject and one session for Semi-blind ICA. Also specify atmost two reference function names

algoType = algo_types(algo_idx);
ica_name = ica_names{algo_idx};

%% Modality. Options are fMRI and EEG
modalityType = 'fMRI';

%% Type of stability analysis
% Options are 1 and 2.
% 1 - Regular Group ICA
% 2 - Group ICA using icasso
% 3 - Group ICA using Minimum spanning tree (MST)
which_analysis = 2;

%% ICASSO options.
% This variable will be used only when which_analysis variable is set to 2.
icasso_opts.sel_mode = 'both'; % Options are 'randinit', 'bootstrap' and 'both'
icasso_opts.num_ica_runs = 20; % Number of times ICA will be run
% Most stable run estimate is based on these settings.
icasso_opts.min_cluster_size = 16; % Minimum cluster size
icasso_opts.max_cluster_size = 20; % Max cluster size. Max is the no. of components

%% Enter TR in seconds. If TRs vary across subjects, TR must be a row vector of length equal to the number of subjects.
TR = 30;

%% Group ica type
% Options are spatial or temporal for fMRI modality. By default, spatial
% ica is run if not specified.
group_ica_type = 'spatial';

%% Parallel info
% enter mode serial or parallel. If parallel, enter number of
% sessions/workers to do job in parallel
parallel_info.mode = 'serial';
parallel_info.num_workers = 4;

%% Group PCA performance settings. Best setting for each option will be selected based on variable MAX_AVAILABLE_RAM in icatb_defaults.m.
% If you have selected option 3 (user specified settings) you need to manually set the PCA options. See manual or other
% templates (icatb/icatb_batch_files/Input_data_subjects_1.m) for more information to set PCA options
%
% Options are:
% 1 - Maximize Performance
% 2 - Less Memory Usage
% 3 - User Specified Settings
perfType = 1;

%% Conserve disk space
% Conserve disk space. Options are:
% 0 - Write all analysis files including intermediate files (PCA, Backreconstruction, scaled component MAT files)
% 1 - Write only necessary files required to resume the analysis. The files written are as follows:
% a. Data reduction files - Only eigen vectors and eigen values are written in the first data reduction step. PCA components are written at the last reduction stage.
% b. Back-reconstruction files - Back-reconstruction files are not written to the disk. The information is computed while doing scaling components step
% c. Scaling components files - Scaling components MAT files are not written when using GIFT or SBM.
% d. Group stats files - Only mean of all data-sets is written.
% 2 - Write all files till the group stats. Cleanup intermediate files at the end of the group stats (PCA, Back-reconstruct, Scaled component MAT files in GIFT). Analysis cannot be
% resumed if there are any changes to the setup parameters. Utilities that work with PCA and Backreconstruction files like Remove components, Percent Variance, etc won't work with this option .
conserve_disk_space = 0;

%% Design matrix selection
% Design matrix (SPM.mat) is used for sorting the components
% temporally (time courses) during display. Design matrix will not be used during the
% analysis stage except for SEMI-BLIND ICA.
% options are ('no', 'same_sub_same_sess', 'same_sub_diff_sess', 'diff_sub_diff_sess')
% 1. 'no' - means no design matrix.
% 2. 'same_sub_same_sess' - same design over subjects and sessions
% 3. 'same_sub_diff_sess' - same design matrix for subjects but different
% over sessions
% 4. 'diff_sub_diff_sess' - means one design matrix per subject.

if algoType == 9
keyword_designMatrix = 'same_sub_same_sess';
else
keyword_designMatrix = 'no';
end

% specify location of design matrix here if you have selected 'same_sub_same_sess' or
% 'same_sub_diff_sess' option for keyword_designMatrix variable
OnedesignMat = fullfile(deriv_dir, 'gift_ica', 'SPM.mat');

%% Specify BIDS info. Data file patterns is read from the bids structure and input_data_file_patterns variable is skipped
% bids_info.root_dir = root_dir;
% bids_info.subjects =;
% bids_info.sessions =;
% bids_info.task =;

%% There are three ways to enter the subject data
% options are 1, 2, 3 or 4
dataSelectionMethod = 4;

%% Method 4
% Input data file pattern for data-sets must be in a cell array. The no. of rows of cell array correspond to no. of subjects
% and columns correspond to sessions. In the below example, there are 3
% subjects and 1 session. If you have multiple sessions, please see
% Input_data_subjects_2.m file.
input_data_file_patterns = {
fullfile(deriv_dir, sub_list{sub_idx}, 'mrs', 'Glu_abs');
};

% Input for design matrices will be used only if you have a design matrix
% for each subject i.e., if you have selected 'diff_sub_diff_sess' for
% variable keyword_designMatrix.
input_design_matrices = {};

% Enter no. of dummy scans to exclude from the group ICA analysis. If you have no dummy scans leave it as 0.
dummy_scans = 0;

%%%%%%%% End for Method 4 %%%%%%%%%%%%

%% Enter directory to put results of analysis
% deriv_dir_old = deriv_dir;
% deriv_dir = fullfile("C:", "Users", "pengh", "gift_ica");
if ~exist(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}), 'dir')
mkdir(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}));
end
if ~exist(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}, ica_name), 'dir')
mkdir(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}, ica_name));
end
if ~exist(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}, ica_name, sub_list{sub_idx}), 'dir')
mkdir(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}, ica_name, sub_list{sub_idx}))
end
outputDir = fullfile(fullfile(deriv_dir, 'gift_ica', prep_folders{preproc_type}, ica_name, sub_list{sub_idx}));
% deriv_dir = deriv_dir_old;
%% Enter Name (Prefix) Of Output Files
prefix = [sub_list{sub_idx} '_' ica_name];

%% Enter location (full file path) of the image file to use as mask
% or use Default mask which is []
maskFile = [];

%% Group PCA Type. Used for analysis on multiple subjects and sessions when 2 data reduction steps are used.
% Options are 'subject specific' and 'grand mean'.
% a. Subject specific - Individual PCA is done on each data-set before group
% PCA is done.
% b. Grand Mean - PCA is done on the mean over all data-sets. Each data-set is
% projected on to the eigen space of the mean before doing group PCA.
%
% NOTE: Grand mean implemented is from FSL Melodic. Make sure that there are
% equal no. of timepoints between data-sets.
%
group_pca_type = 'subject specific';

%% Back reconstruction type. Options are 1 and 2
% 1 - Regular
% 2 - Spatial-temporal Regression
% 3 - GICA3
% 4 - GICA
% 5 - GIG-ICA
backReconType = 4;

%% Maximum reduction steps you can select is 2
% You have the option to select one data-reduction or 2 data reduction
% steps when spatial ica is used. For temporal ica, only one data-reduction
% is done.
numReductionSteps = 2;

%% Batch Estimation. If 1 is specified then estimation of
% the components takes place and the corresponding PC numbers are associated
% Options are 1 or 0
doEstimation = 0;

%% MDL Estimation options. This variable will be used only if doEstimation is set to 1.
% Options are 'mean', 'median' and 'max' for each reduction step. The length of cell is equal to
% the no. of data reductions used.
estimation_opts.PC1 = 'max';
estimation_opts.PC2 = 'mean';

%% Number of pc to reduce each subject down to at each reduction step
% The number of independent components the will be extracted is the same as
% the number of principal components after the final data reduction step.
numOfPC1 = sub_n_components{sub_idx};
numOfPC2 = sub_n_components{sub_idx};

%% Scale the Results. Options are 0, 1, 2
% 0 - Don't scale
% 1 - Scale to Percent signal change
% 2 - Scale to Z scores
scaleType = 2;

%% Report generator (fmri and smri only)
% display_results.formatName = 'html';
% display_results.slices_in_mm = (-40:4:72);
% display_results.convert_to_zscores = 'yes';
% display_results.threshold = 1.0;
% display_results.image_values = 'positive and negative';
% display_results.slice_plane = 'axial';

% display_results.anatomical_file = 'C:\Users\srrac\Desktop\GroupICATv4.0c\icatb\icatb_templates\ch2bet_3x3x3.nii';

% %Network names and components are used in the plots (only fmri). If you are using
% %moo-icar or constrained ica (spatial), you can specify network names and
% %components within each network. Below is an example from neuromark
% %template labels
% display_results.network_summary_opts.comp_network_names = { 'SC', (1:5);
% 'AU', (6:7);
% 'SM', (8:16);
% 'VI', (17:25);
% 'CC', (26:42);
% 'DM', (43:49);
% 'CB', (50:53)};
% display_results.network_summary_opts.outputDir = fullfile(outputDir, 'network_summary');
% display_results.network_summary_opts.prefix = [prefix, '_network_summary'];
% display_results.network_summary_opts.structFile = display_results.anatomical_file;
% display_results.network_summary_opts.image_values = display_results.image_values;
% display_results.network_summary_opts.threshold = display_results.threshold;
% display_results.network_summary_opts.convert_to_z = display_results.convert_to_zscores;
% %some more network summary options
%display_results.network_summary_opts.conn_threshold = 0.2;
%display_results.network_summary_opts.fnc_colorbar_label = 'Corr';
% options are 'slices' and 'render'
%display_results.network_summary_opts.display_type = 'slices';
%display_results.network_summary_opts.slice_plane = 'axial';
% colormap of the correlations
%display_results.network_summary_opts.cmap = jet(64);
% CLIM - range of the data values in [min_value, max_value] format
%display_results.network_summary_opts.CLIM=CLIM;

%% ICA Options - Name by value pairs in a cell array. Options will vary depending on the algorithm. See icatb_icaOptions for more details. Some options are shown below.
% Infomax - {'posact', 'off', 'sphering', 'on', 'bias', 'on', 'extended', 0}
% FastICA - {'approach', 'symm', 'g', 'tanh', 'stabilization', 'on'}

icaOptions = {'posact', 'off', 'sphering', 'on', 'bias', 'on', 'extended', 0};

%% Specify atmost two reference function names if you select Semi-blind ICA algorithm.
% Reference function names can be acessed by loading SPM.mat in MATLAB and accessing
% structure SPM.xX.name.
% refFunNames = {'Sn(1) rightbf(1)', 'Sn(1) leftbf(1)'};

%% Specify spatial reference files for constrained ICA (spatial) or moo-icar
% refFiles = {which('ref_default_mode.nii'), which('ref_left_visuomotor.nii'), which('ref_right_visuomotor.nii')};
`

This is one of the MATLAB crash reports:

MATLAB Log File: /home/maomao/matlab_crash_dump.8080-1

---

MATLAB Log File

---

      Segmentation violation detected at 2025-06-01 00:02:50 +0200

---

Configuration:
Crash Decoding : Disabled - No sandbox or build area path
Crash Mode : continue (default)
Default Encoding : UTF-8
Deployed : false
GNU C Library : 2.35 stable
Graphics Driver : Uninitialized hardware
Graphics card 1 : 0x1414 ( 0x1414 ) 0x8e Version 2.0.3.0 (0-0-0)
Graphics card 2 : 0x1414 ( 0x1414 ) 0x8e Version 2.0.3.0 (0-0-0)
Java Version : Java 1.8.0_202-b08 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode
MATLAB Architecture : glnxa64
MATLAB Entitlement ID : 1376434
MATLAB Root : /home/maomao/MATLAB
MATLAB Version : 25.1.0.2943329 (R2025a)
OpenGL : hardware
Operating System : Ubuntu 22.04.5 LTS
Process ID : 8080
Processor ID : x86 Family 6 Model 183 Stepping 1, GenuineIntel
Session Key : 6lqo3sm0uep77dt5heo3n4qm
Window System : Microsoft Corporation (12010000), display :0

Fault Count: 1

Abnormal termination:
Segmentation violation

Current Thread: 'MCR 0 interpret' id 140595769960000

Register State (from fault):
RAX = 0000000000000000 RBX = 00007fdba58e1830
RCX = 0000000000000001 RDX = 00007fdd5c3a4cfc
RSP = 00007fdf00ec88f0 RBP = 00007fdbaece4290
RSI = 00007fdea40008e0 RDI = 0000000000000007

R8 = 00007fdc5c268d00 R9 = 0d0000000000000b
R10 = 0000000029303000 R11 = a0a5c38644a7ff87
R12 = 00007fdbaece4290 R13 = 00007fdbaece4290
R14 = 00007fdba58e1830 R15 = 00007fdd5c2e4b30

RIP = 00007fdba58e1830 EFL = 0000000000010246

CS = 0033 FS = 0000 GS = 0000

Stack Trace (from fault):
[ 0] 0x00007fdba58e1830 +00000000

[PenghuiDu]

This is another common error, might be relevant:

Launch Icasso visualization supposing 20 estimate-clusters.
Show demixing matrix rows.
Launch Icasso visualization supposing 20 estimate-clusters.
Show IC source estimates (default), reduce number of lines
Collect results.

......................................
Group ICA Error Information:

Index in position 2 exceeds array bounds. Index must not exceed 352.

Error in ==> getStableRunEstimates at 728
Error in ==> icatb_calculateICA at 382
Error in ==> icatb_runAnalysis at 431
Error in ==> icatb_batch_file_run at 73
Error in ==> fmrsi_gift_ica_batch_analysis at 90
......................................
Error using icatb_displayErrorMsg (line 23)

Error in icatb_runAnalysis (line 559)
icatb_displayErrorMsg;
^^^^^^^^^^^^^^^^^^^^^
Error in icatb_batch_file_run (line 73)
sesInfo = icatb_runAnalysis(sesInfo, 1);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in fmrsi_gift_ica_batch_analysis (line 90)
icatb_batch_file_run(fullfile(work_dir, 'code', 'mrs', 'GIFT_ICA', 'fmrsi_gift_ica_pref.m'))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^