Re: [dsi-studio] Inquiry regarding AutoTrack parameter optimization for Lobe-Specific Corticopontine (Frontal, Parietal, Occipital) and DRTT Pathways in the Sudmex CONN Dataset

[Frank Yeh]

Dear Ayşegül,

Thank you for the detailed description.

If increasing the angular threshold to 75° or 90° generates Z-shaped or hook-like streamlines that merge into neighboring frontopontine pathways, I would treat those results as false positives rather than an improvement. A higher angular threshold can recover more streamlines, but it can also allow anatomically implausible turns, especially near the brainstem and cerebellar pathways.

For AutoTrack, I would suggest keeping the analysis conservative:

1. Use the same QSDR reconstruction and tracking settings for all subjects.
2. First check whole-brain tracking and registration quality visually.
3. Do not tune parameters separately for each tract or each subject just to force a pathway to appear, because this can introduce bias.
4. You may run a predefined sensitivity analysis with a small number of tolerance/seed settings, but the final rule should be fixed before comparing groups.

For the parietopontine, occipitopontine, and DRTT pathways, failure in many subjects usually means the data quality, resolution, b-value/directions, or tract-recognition atlas is not sufficient for reliable automated reconstruction. DRTT is especially challenging because it is small and passes through complex crossing regions near the brainstem.

Loading a different tract atlas may help only if the atlas was specifically constructed and validated for these pathways in the same template space. Otherwise, it may not solve the problem and could introduce additional uncertainty.

If frontopontine tracts are robust but the posterior corticopontine and DRTT pathways are not, I think it is methodologically safer to report the robust pathways and state that the other pathways could not be reliably reconstructed with the fully automated pipeline and available acquisition. I would avoid reporting parameter-forced tracts if they show clear Z-shape or hook artifacts.

Best regards,
Frank

On Fri, 3 Jul 2026 20:39:15 +0300, “Ayşegül Ayran” aysegul....@gmail.com wrote:

Dear Dr. Yeh,

First of all, thank you for developing DSI Studio and the fantastic AutoTrack framework, which has been immensely helpful for our neuroimaging research.

I am currently analyzing the open-access Sudmex CONN dataset (a multi-shell diffusion MRI dataset focusing on Cocaine Use Disorder) for a study focusing on cerebro-cerebellar connectivity. The data was reconstructed using QSDR to align with the MNI template space as required by the AutoTrack pipeline.

In this study, we are strictly utilizing the automated AutoTrack feature, focusing on the Corticopontine pathways (specifically analyzing the sub-pathways originating from the Frontal, Parietal, and Occipital regions to investigate lobe-specific profiles) along with the Dentatorubrothalamic Tract (DRTT). To maintain strict methodological automation and avoid observer bias, we are not using any manual ROI drawing.

When running AutoTrack on several subjects from this dataset using the default Angular Threshold = 0 (random/automatic) configuration, we observed a very specific and challenging behavior: Frontopontine Tracts track beautifully and robustly, yielding around 46,000 to 48,000 streamlines per hemisphere.

Parietopontine and Occipitopontine Tracts completely fail to form robust bundles. Instead of proper pathways, they yield either exactly 0 or just a few fragmented streamlines (ranging randomly from 20, 30, 40 up to ~200 tracks) that lack structural integrity.

Dentatorubrothalamic Tracts (DRTT) also yield 0 streamlines in these same individuals.To investigate this further within the automated framework, I performed an experimental check by changing the explicit angular threshold to 75° and 90°. While this increases the streamline counts slightly, the generated tracks form sharp, artifactual “Z-shapes” or “hooks” right above the lower brainstem boundary, eventually merging into and hijacking the neighboring frontopontine pathway.

We would highly appreciate your expert guidance on the following questions regarding the AutoTrack pipeline:

How can we optimize the internal AutoTrack parameters (such as adjusting tracking thresholds, track-to-voxel ratio, or tolerance) within the interface to successfully map and measure the Parietopontine and Occipitopontine pathways when they fail under default settings, without producing false-positive “Z-shape” artifacts?

Would loading a different tract atlas into the AutoTrack framework help resolve this? For the Dentatorubrothalamic Tract (DRTT), are there any specific AutoTrack-internal settings or parameter tunings you recommend to help the algorithm successfully recognize this ascending pathway when it yields 0 streamlines under default configurations?

If the automated AutoTrack pipeline cannot reliably resolve these specific posterior/ascending pathways under default or optimized parameters, would you agree that it is methodologically safer to report only our robust Frontopontine tracts for these subjects and declare the other pathways as missing/untrackable due to acquisition limits in our publication?

In short, how can I successfully reconstruct and quantify these specific Parietopontine, Occipitopontine, and Dentatorubrothalamic (DRTT) pathways within the automated framework? Our ultimate goal is to be able to measure and analyze them with the same level of robustness, stability, and reliability as we currently achieve with the Frontopontine tracts.

Thank you very much for your time, guidance, and invaluable support.

Best regards,

Ayşegül Ayran

–

You received this message because you are subscribed to the Google Groups “DSI Studio” group.

To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/CAJV5AwfvaKjgxekuKyRbgKSKc0%2ByZdL5ZmC%2BJY1GAJU9gNK5QA%40mail.gmail.com.