The video at this link gives a glimpse into one future of stat analysis of massive time series matrices . Here we see:
“Integrated regression and/or ML/DL with relational time patterns of vectors as features (predictors).”
Diring sepsis of Covid-19 will we see the relational perturbation vectors grow in number of vectors and/or or duration of vectors or they are replaced by recovery vectors over time. This produces a time "image"which imay be recognized.
So the relational time patterns of perturbation and recovery vectors which include treatments all n relation to demographics, comorbiditie, comprise the time image predictive of future clinical state.
Contrast this to the traditional selection of threshold scores responsive to a few select time series of the matrix (eg SOFA).