diff --git a/README.md b/README.md
index 2943921657522a2691e2c2bd742242876993cda7..354c15a0a67e8f648c1f496f816edb353504ca6b 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# tvb-make
+# pipeline
This is a workflow for preparing and using TVB brain network models, comprised of
three main components
@@ -59,9 +59,27 @@ outputs.
### Targets
-- `fs-recon` - FreeSurfer reconstruction, `recon-all -all ...`
-- `resamp-anat` - Lower resolution cortical surfaces & annotations
-- `conn` - Connectivity files
+- `fs-recon`: FreeSurfer reconstruction. Consists mainly of running `recon-all -all`.
+ Uses `T1`.
+
+- `resamp-anat`: Lower resolution cortical surfaces & annotations
+ Uses `T1`.
+
+- `conn`: Connectivity matrices in text format.
+ Uses `T1` and `DWI`.
+
+- `tvb`: TVB zipfile, cortical and subcortical surfaces in TVB formats, region mappings.
+ Uses `T1` and `DWI`.
+
+- `elec`: Positions of the contacts of depth electrodes and gain matrices.
+ Uses `T1`, `DWI`, `ELEC`, and either `ELEC_ENDPOINTS` or `ELEC_POS_GARDEL`.
+
+- `seeg`: Conversion of SEEG recordings to FIF format and plotting the recordings.
+ Uses `SEEGRECDIR`, `XLSX` and everything that `elec` uses.
+
+- `ez`: Extraction of the epileptogenic zone from the patient Excel file.
+ Uses `XLSX` and everything that `elec` uses.
+
_TODO_ more details & help on this
@@ -107,13 +125,21 @@ environment variable.
### Marseille Cluster
-The `cluster/run` script assists in running the pipeline on the Marseille
-cluster through two modes. First, invoke with typical arguments
+For quick introduction look at the basic [step-by-step tutorial](doc/TutorialCluster.md).
+
+There are two options for running the pipeline on the cluster: non-interactive and interactive.
+For running the full pipeline, non-interactive mode is recommended due to the large time requirements.
+For small updates and testing the interactive mode might be more suitable.
+
+#### Non-interactive mode
+
+In the non-interactive regime, you prepare the data and submit the job(s), and the scheduler takes cares of the execution.
+The `cluster/run` script assists in running the pipeline on the Marseille cluster through two modes.
+ First, invoke with typical arguments
```bash
-cluster/run SUBJECTS_DIR=fs SUBJECT=foo T1=data/T1.nii.gz fs-recon
+<PIPELINE_DIR>/cluster/run SUBJECTS_DIR=fs SUBJECT=foo T1=data/T1.nii.gz fs-recon
```
-for a single run in a single OAR job, or for many subjects,
-create a file `params.txt` with multiple lines of arguments, e.g.
+for a single run in a single SLURM job. If you have many subjects, create a file `params.txt` with multiple lines of arguments, e.g.
```
SUBJECTS_DIR=fs SUBJECT=foo T1=data/T1.nii.gz fs-recon
SUBJECTS_DIR=fs SUBJECT=bar T1=data/T2.nii.gz fs-recon conn
@@ -121,15 +147,41 @@ SUBJECTS_DIR=fs SUBJECT=baz T1=data/T3.nii.gz conn
```
then
```
-cluster/run params.txt
+<PIPELINE_DIR>/cluster/run params.txt
+```
+Each line will result in the pipeline running a SLURM job for every line. You can comment out a line if you prepend it with a `#` sign,
+```
+ # SUBJECTS_DIR=fs SUBJECT=foo T1=data/T1.nii.gz fs-recon
```
-Each line will result in the pipeline running once for the arguments
-on a given line, and an OAR job.
NB You need to provide a custom, valid FreeSurfer `SUBJECTS_DIR`,
since the default directories on the cluster (`/soft/freesurfer*/subjects`)
are not writeable by users.
+#### Interactive mode
+
+First, request a computational node in the interactive mode
+```
+srun --pty bash
+```
+which should give you the interactive node if there is one available.
+
+If you need to run the reconstruction and tractography in the interactive mode
+(although that is discouraged), you need to request full node with enough memory:
+```
+srun -N 1 -n 1 --exclusive --mem=60G --pty bash
+```
+
+Then setup your working environment by loading the environment file,
+```
+source <PIPELINE_DIR>/cluster/env
+```
+and run `make` by hand:
+```
+make -f <PIPELINE_DIR>/Makefile SUBJECTS_DIR=fs SUBJECT=foo T1=data/T1.nii.gz fs-recon
+```
+
+
## Special Cases
### JPEG encoded images
@@ -181,4 +233,4 @@ which generate or use files in the `stan` subfolder of the subjects' folder
- `$(sd)/stan/{model_name}.samp.pkl` - posterior samples found during fit
- `$(sd)/stan/{model_name}.png` - visualization produced by `stan/{model_name}.vis.py`
-See the [`stan`](stan) folder for an example, to be completed.
\ No newline at end of file
+See the [`stan`](stan) folder for an example, to be completed.
diff --git a/doc/TutorialCluster.md b/doc/TutorialCluster.md
new file mode 100644
index 0000000000000000000000000000000000000000..9cb4afa29710449490c121dae228c0eb755426d9
--- /dev/null
+++ b/doc/TutorialCluster.md
@@ -0,0 +1,101 @@
+
+
+
+# Running the pipeline on INS cluster
+
+
+Brief tutorial on how to run the reconstruction pipeline on the INS cluster.
+
+## Setting up the environment (do once)
+
+Clone the pipeline repository. Note that you need your SSH key generated on the cluster to be added
+in the Gitlab interface to be able to clone the repo from the cluster.
+```
+cd ~/soft # Or wherever you want the code to be
+git clone git@gitlab.thevirtualbrain.org:tvb/pipeline.git
+```
+
+## Preparing the data
+
+By default, the pipeline uses following data structure:
+```
+data/SUBJECT-1/
+ SUBJECT-2/
+ SUBJECT-3/
+ ...
+fs/SUBJECT-1/
+ SUBJECT-2/
+ SUBJECT-3/
+ ...
+```
+where `data/` contains the raw data, and `fs/` contains the processed results. You should prepare
+the contents of the `data/` directory; the contents of the `fs/` directory is filled by the pipeline.
+If needed, you can change the the names of the raw data directory by setting the make variable `DATA`
+(see below on how to), and the `fs/` directory by the variable `SUBJECTS_DIR`.
+
+
+First, let's create this main directory structure in `~/reconstruction`
+```
+cd
+mkdir -p reconstruction/data reconstruction/fs
+```
+
+
+Then start with a single patient `SUBJECT-1`.
+For the basic TVB dataset, you need at least T1 and DWI scans. Then place the T1 and DWI scans under `t1/`
+and `dwi/` directories under the patient directory:
+```
+data/SUBJECT-1/dwi/
+ t1/
+```
+
+
+## Running the pipeline
+
+You need to create a file containing the subject specification. In your working directory, create
+a file `params.txt` and insert a single line inside:
+```
+SUBJECT=SUBJECT-1 T1=data/SUBJECT-1/t1/t1.nii.gz DWI=data/SUBJECT-1/dwi/dwi.nii.gz tvb
+```
+Now what this means? Every line specifies a single subject, so here we have specified a single
+subject named `SUBJECT-1`.
+By setting the variables `T1` and `DWI` we have told the pipeline where the raw data are.
+If you have the raw data in DICOM format (many .dcm files in `t1` and `dwi`
+directories) and not single files, simply point to the directories:
+```
+SUBJECT=SUBJECT-1 T1=data/SUBJECT-1/t1/ DWI=data/SUBJECT-1/dwi/ tvb
+```
+
+Last keyword on this line is the *target* of the pipeline. In this case, `tvb` stands for the
+TVB data set with connectomes and surfaces. Other targets that may be useful are `fs-recon` for
+the FreeSurfer reconstruction only, `elec` for the depth electrode positions, or `seeg` for SEEG
+recordings.
+
+
+The pipeline job is submitted simply by running the following command:
+
+```
+~/soft/pipeline/cluster/run params.txt
+```
+The output of the command should show that for every line in the `params.txt` (not commented out)
+a Slurm job was submitted.
+
+## Examining the results
+
+The status of the Slurm jobs on the cluster can be checked by
+```
+squeue -u USERNAME
+```
+If you have submitted a job and it is finished after only a few seconds, something probably
+went wrong. Have a look at the logs in `fs/_logs/SUBJECT-1.*.stdout` and
+ `fs/_logs/SUBJECT-1.*.stderr`.
+
+After the job has ended, examine the logs and the created subject directory
+`fs/SUBJECT-1/`, especially the `tvb` subdirectory, where your desired data (TVB zipfiles,
+ surfaces, region mappings) should be.
+
+
+
+
+
+