Pretherapeutic Identification of High Risk AML Patients
fig.4: Reclassification of the Learning Set
- Display of the disease classification masks for ten sequential
5-year survivor and 5-year nonsurvivor AML patients
- The masks
for the reference patients ( > 5-year,
> 2 year survival) typically contain a majority of (0)
characters because the parameter values are located between the lower and upper
percentile thresholds (10/90, 15/85, 20/80, 25/75 or 30/70%) of the respective value distributions.
As an example, in case of the 10/90% percentile pair, 80% of the values are located between
both percentile thresholds that is they are (0), 10% are below (-) the lower and 10% above (+) the
upper percentile thresholds.
- Unknown patients are classified according to highest positional
coincidence of their patient classification mask with any one of
the disease classification masks.
- The degree of coincidence between the patient mask
and the best fitting disease mask is expressed by
the disease coincidence factor. The coincidence factor
is 1.00 for patients #100073 and #080055 (arrows) with four
triple matrix characters (0) and three characters (-)
despite the fact that not all triple matrix characters are (0).
This is because triple matrix characters not occuring in the
disease classification mask of 5-year nonsurvivors count
as positional coincidences for 5-year survivors.
- This results in 19 correctly classified patient at 1
- The patient disease masks labelled a) and b) are
repetitive. The systematic analysis of disease masks
may permit to define leukemia subgroups characterized by certain
preferential molecular CD expression patterns.
last update: Jun 03,2022
first display: Mar 31,2003