Handle adni missing session numbers

[omar-rifai]

Currently, when converting data with adni-to-bids, we have some inconsistencies when handling sesssions:

1. In some clinical data files, we have different records with the same session ID:

• E.g in ECOGPT.csv, for subject '003_S_4354' we have two records with different data for session m84)
• Digging further, we see that (the now obsolete) VISCODE field has two different values y1and y2. However the VISCODE2 field which is used has (incorrectly) the same value.
• The exam date gives use the right information since the records are 6 months apart
• The second record should have had session m90 (which currently doesn't exist in ADNIMERGE.csv)

Currently the workaround consists in only taking the latest data record for the existing session but a better solution would be to introduce the ses-m90 (even if it doesn't exist in ADNIMERGE)

2. Some records have a missing session number:
   • E.g in UWNPSYCHSUM_03_26_20.csv for subject 003_S_4555, we have a record with value nv for both VISCODEand VISCODE2

• According to a the adni forum

"For the ADNI-GO/2 file [...] “nv” usually refers to a scan in which the EXAMDATE does not match information that we have about visit dates or metadata for a scan being done, so it isn’t assigned a specific visit."

Currently these records are removed from resulting sessions.tsv files to prevent inconsistency with clinicaDL. A better solution would be to find the exam_date(if it is properly informed) and add the related session name accordingly.