Patient Prognosis after Melanoma Surgery
- The print output of the CLASSIF1 learning procedure MELANOMA comprises the confusion matrix for the classification of the learning set and the unknown test set patients, (fig.1), the reclassification of the learning set (fig.2), the disease classification masks (fig.3), the parameter list of the database (fig.4), the means, SEMs and percentiles of the discriminatory parameters of the disease classification masks (fig.5) and the prospective classification of unknown test set patients (fig.6).
- Surviving and non surviving melanoma patients are
prognosticated with predictive values of
80.3% and 79.8% from the parameters melanoma
thickness, mean of thickness+infiltration depth (Clark level) and
% S-phase cells
- The reclassification of the learning set (fig.2) shows that the sample classification masks do not exhibit systematic data pattern differences with the increasing number of patients i.e. the clinical and flow cytometric measurements were of stable quality over the collection period of the samples which was > 10 years.
- The reclassification table shows furthermore that correct classification may be obtained despite non total positional coincidence of the: +, - and 0 characters with the patient classification masks and the disease classification masks as evidenced by mask coincidence factors < 1.00 (fig.2).
- The unknown test set patients (fig.6) classify similarly (fig.1) as the learning set patients (fig.2) indicating robustness of the CLASSIF1 classification process towards the prospective classification of unknown persons.
- The selected parameters of the disease classification masks (fig.3) differ statistically significantly between surviving and non surviving melanoma patients (fig.5). The 3 selected parameters constitute 50.0% of the totally available information of 6 parameters (fig.4).
|© 2023 G.Valet|