- The triple matrix converted data can classified, once the disease
classification masks ("disease signature") are established. They consist of the most frequent
triple matrix character in each data column and classification category of
the learning set.
- The most frequent triple matrix character for the value distribution
of reference patients
(fig.3a)
is (0) and (+) for the value distribution of the abnormal group
(fig.3c).
- The disease classification mask for reference patients contains
therefore (0) and for the abnormal group (+). The (-) triple matrix character
does not occur in the abnormal group and is additionally assigned to the
disease classification mask of reference patients.
- Reclassification of the learning set value distributions
(fig.3a/3c)
according to the above disease classification masks provides the
correct classification of 90% (9 of 10) of the reference patients and
of 60% (6 of 10) of the abnormal group patients.
- The discrimination potential of triple matrix data pattern classification
increases with the number of discriminatory parameters in a given database
(fig.9).
It can typically reach classification accuracies of >95% or 99%
which is of particular interest for
predictive (individualized/personalized)
medicine.
