May be worth adding a style checker to the CI. For example, https://github.com/florianschanda/miss_hit looks interesting and has a template for Github Actions here: https://florianschanda.github.io/miss_hit/configuration.html

Originally posted by @koenhelwegen in #19 (comment)

As in the paper, CATO framework was tested on the HCP dataset, i wonder if you could provide some json templates for some most-used MRI dataset (HCP dataset for example).
In detail, there was just default settings in the configuration assistant, and i wonder if there could have some json template for the mose-used dataset.

i will appreciate if you could consider my suggestion.

When applying a bandpass filter in the functional pipeline, we sometimes observe artifacts in the beginning and/or end of the timeseries. @SiemondeLange and I looked into this and noticed that matlabs filtfilt function matches the initial and final few timepoints of the filtered signal to the unfiltered signal. For our use case this is undesirable, as the amplitude of the high-frequency signal that is removed during filtering is higher than the mid-frequency signal that is retained; matching the initial/final signal causes the observed artifacts.

To solve this, we will avoid the use of filtfilt and apply the filter directly.

Currently, CATO only supports deterministic fiber tracking. While, deterministic tracking is widely used in the field of connectomics and has a high specificity (beneficial in connectome studies), there is a strong argument for providing probabilistic fiber tracking (PFT) to CATO.

PFT allows for the modelling of uncertainty in the estimated fiber pathways, providing a more detailed understanding of the underlying white matter connectivity. Moreover, PFT has been shown to have a greater sensitivity (and its specificity can be improved with streamline filtering methods).

The implementation of PFT in CATO could be approached in two possible ways. The first approach would be to integrate PFT directly into the fiber_reconstruction pipeline step, making the implementation as accessible to users as possible. Alternatively, PFT could be added as an add-on, providing more flexibility to use external software (e.g. MRtrix3) for the probabilistic fiber tracking.

As a first step, I suggest to make an inventory of possible PFT algorithms and implementations that we can use to decide on the specific PFT implementation in CATO.