Bug Report: Errors Converting Siemens Log Files to BIDS ('Index exceeds...' & 'Size mismatch')

[angelaaseo]

Hello PhysIO Development Team,

First, thank you for developing and maintaining this incredibly useful toolbox.

To give an overview, our physiological signals are recorded in one continuous run, while fMRI sequences are parsed into several runs. Therefore, I am attempting to use the PhysIO toolbox to both time-align the physio & MRI modalities, as well as to convert these to BIDS format. However, I am running into a persistent issue with the BIDS conversion, which appears to be related to the file parsing stage, with different errors occurring depending on the input file type.

Setup
• Scanner & sequence: Siemens 3 T (VB17), multi‑echo EPI (3 echoes/TR), multiband = 2
• ECG, Pulse-oximetry, Respiration data: Manual recording via Telnet MPCU (continuous across several fMRI blocks)
• Software: TAPAS Version 6.0.1, PhysIO Version 9.0, MATLAB R2020b, SPM12 r6685, Windows 10 (Virtual Desktop)

Context:
I initially faced issues related to MATLAB path conflicts between a new TAPAS installation and an older, bundled PhysIO version within a server-based SPM installation. To eliminate these environmental factors, I have since created a clean, local setup by installing the latest versions of both SPM12 and the TAPAS toolbox in my local Documents\MATLAB folder. The errors described below are from this clean, local setup.

Problem Description
When I run a script to perform BIDS conversion using tapas_physio, the process fails with one of two errors, depending on whether I use a .puls or .ecg file for the cardiac data.

Scenario 1: Using .puls and .resp files

The job fails with an "Index exceeds the number of array elements" error. This suggests the file parser is reading a line but failing to extract any data, resulting in an empty array.

Error Log:

29-Sep-2025 11:30:19 - Failed  'TAPAS PhysIO Toolbox'
Index exceeds the number of array elements. Index must not exceed 0.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_siemens_line2table.m", function "tapas_physio_siemens_line2table" at line 71.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles_siemens.m", function "tapas_physio_read_physlogfiles_siemens" at line 197.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles.m", function "tapas_physio_read_physlogfiles" at line 92.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_main_create_regressors.m", function "tapas_physio_main_create_regressors" at line 138.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_cfg_matlabbatch.m", function "run_physio" at line 1666.

Scenario 2: Using .ecg and .resp files

When I swap the .puls file for the corresponding .ecg file (from the same session), the error changes to a size mismatch error. This "off-by-one" error suggests a subtle difference in how the .ecg file is being parsed compared to the .puls file.

Error Log:

29-Sep-2025 13:34:50 - Failed  'TAPAS PhysIO Toolbox'
Unable to perform assignment because the size of the left side is 918491-by-1 and the size of the right side is 918490-by-1.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_siemens_line2table.m", function "tapas_physio_siemens_line2table" at line 165.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles_siemens.m", function "tapas_physio_read_physlogfiles_siemens" at line 197.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles.m", function "tapas_physio_read_physlogfiles" at line 92.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_main_create_regressors.m", function "tapas_physio_main_create_regressors" at line 138.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_cfg_matlabbatch.m", function "run_physio" at line 1666.

The following modules did not run:
Failed: TAPAS PhysIO Toolbox

Warning: An error occurred during processing. 
Error using MATLABbatch system
Job execution failed. The full log of this run can be found in MATLAB command window, starting
with the lines (look for the line showing the exact #job as displayed in this error message)

Summary & Question
Both errors point towards an issue within the Siemens log file reader. My current guess is that my specific log files may have a subtle formatting anomaly that the parser is not robust to.

Could you please provide any guidance on:
Whether these are known issues with certain Siemens scanner log files (e.g., a software version issue)?
What specific formatting issues should I look for within the raw .puls, .resp, and .ecg files?
Would there be any known work-arounds to debug this file-reading process to get the BIDS conversion working?

Thank you for your time and assistance.
Best,
Angela Seo

In case, it is helpful, I am attaching the following:

1. A screenshot of the first and last lines of the .puls file opened as txt

2. A screenshot of the first and last lines of the .ecg file opened as txt

[mrikasper]

Dear Angela,

Thank you for using the PhysIO Toolbox.

It seems to be a subtle issue with your logfiles indeed, because some dimensions do not seem to match when reformatting the data stream into channel columns in line 165 of tapas_physio_siemens_line2table.

Would you be able to share the script/matlabbatch you used to run PhysIO, and all the physiological logfiles of one subject (.puls., .ecg, etc), so that I can reproduce the behavior? You can send the data to my email, mentioned in the README of PhysIO.

In the meantime, do you need both the ECG and the .puls data? Does it work if you only specify one as cardiac recording file?

Looking forward to your response and

All the best,
Lars

[mrikasper]

Dear Angela @angelaaseo,

Thank you for sending along the logfiles and sorry for the long delay. Johanna @likeajumprope and I looked at the files, and they seem a bit idiosyncratic, i.e., certain header information is not written out as expected:

• Have a look at your .puls file compared to our example one in examples/Siemens_VB/PPU_3T_Sync_First:
  • your first line: 1 2 40 280 1604 1622 1724 1946 2246 2564
  • example first line 1 2 40 280 5002 LOGVERSION_PULS 1 6002 1944 1959 1974 2003 2018 2018
  • This means the 5002 and 6002 flags are missing which typically indicate the end of the "header numbers" (usually the first 4 or 5 in the first line), and start of the actual pulse trace time series, respectively
  • PhysIO crashes, because it cannot find the 6002 flag that would indicate the start of the trace in the file.
• We can guess that only the first 4 numbers in your file are the header, followed directly by the trace, but this ad-hoc solution only makes sense, if you see such a pattern for other participants/runs as well
• One way to identify the first few numbers as header and not trace values is looking at the range reported at the end of the .puls file:

  PULS Min Max Avg StdDiff: 516 2030 700 4
  RESP Min Max Avg StdDiff: 2460 7120 3427 107
  

  But they also don't seem to correspond to the correct range, so it's hard to interpret this information.
• Similarly, in your .ecg file, the number of data points (after 6002, because here the header seems correct) is odd, even though it should be even (2 ECG channels). This means, at least one data point wasn't written away.

Our overall conclusion was that maybe the telnet device that you start with the ideaCmdTool skipped writing the header or some data in between. We may find a workaround for that, but I wanted to first ask how many of your runs/sessions/participants are affected by this behavior, and whether you recall anything odd happening at the scanner during these recordings.

Thank you for your input,
Johanna and Lars

[mrikasper]

Dear Angela @angelaaseo ,

Happy New Year! I just wanted to bubble up this issue and ask whether you can provide any input on the questions we had regarding the affected data, in particular

how many of your runs/sessions/participants are affected by this behavior, and whether you recall anything odd happening at the scanner during these recordings

Thank you and

All the best,
Lars

[angelaaseo]

Dear Lars @mrikasper ,
Thank you for reviewing our data and for identifying the potential issues.
It seems like the issue regarding the missing headers affects all of our PPG and respiration files (all runs/sessions/participants). No .puls or .resp files contain the full header line (including 5002 and 6002 markers) and the physio trace begins immediately after "1 2 40 280".

On the other hand, we recently noticed another unfortunate issue, which is that many of our ECG recordings are cut short at varying times (before the end of the fMRI run)- including the example run I sent via email. I suspected this may explain the ECG error of mismatched arrays.
However, I tried running the batch using a "good" ECG session (one that did not cut short), but still encountered the same error.

`Unable to perform assignment because the size of the left side is 1204630-by-1 and the size of the right side is 1204629-by-1.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_siemens_line2table.m", function "tapas_physio_siemens_line2table" at line 165.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles_siemens.m", function "tapas_physio_read_physlogfiles_siemens" at line 197.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\readin\tapas_physio_read_physlogfiles.m", function "tapas_physio_read_physlogfiles" at line 92.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_main_create_regressors.m", function "tapas_physio_main_create_regressors" at line 138.
In file "C:\Users\seo\Documents\MATLAB\spm\toolbox\PhysIO\tapas_physio_cfg_matlabbatch.m", function "run_physio" at line 1666.

The following modules did not run:
Failed: TAPAS PhysIO Toolbox`

If a work-around exists for either cases, this would be very helpful!
Best,
Angela

[mrikasper]

Dear Angela,

Thank you for following up.

1. If all of your files start with 1 2 40 280, we can certainly create an extra case in our read-in to ignore the typically expected header. I am away next week, but can try to implement it after that.
2. Can you please send me the example files with the ECG not being cut short that triggered the error in your last message? It would be great to check the new implementation with this dataset as well.
3. It's interesting that you mention the files got cut short. How did you use the ideaCmdTool? Typically, if you only use some of the recordings (like PPU and resp, but not ECG), you have to tell the telnet tool explicitly to not record the other modalities, otherwise it stops recording for all of them after 5 minutes. Are the files always 5 minutes long?

Looking forward to your response,
Johanna and Lars