! - file: classi12.html, first display: Feb 16, 2005 ->
CLASSIF1 Data Pattern Classification
- The disease classification mask ("disease signature")
for the reference group of individuals (normals) contains
typically a sequence of (0) characters because the majority (70%)
of the parameter values are located between the two percentiles thresholds
15% and 85%, 15% below (-) the lower percentile and 15% above (+) the
- Unknown test set patients are classified according to the highest positional coincidence of the patient classification mask ("patient signature") with any of the disease classification masks .
- The degree of overall coincidence between the patient mask and the best fitting disease mask is expressed as mask coincidence factor. The coincidence factor is 1.00 for risk patients #137,139,141,142,143,148 but also for non risk (normal) patients #102,108,109 (arrows <= ) despite the fact that not all triple matrix characters are (0). Fully coincident infarct risk patients have all discriminatory parameters of the disease mask increased (+). Patients with normal (0) or diminished (-) discriminatory parameter values belong therefore to the non risk (normal) patients. This means that parameter values (-) are counted as hit for non risk (normal) patients, explaining the coincidence factor of 1.00 for patients #102,108,109.
- The triple matrix patterns of the first 10 patients of the normal individuals and the myocardial risk patients are displayed in the above graph. The missing patients #101,105,110,140,145 belong to the unknown test set patients, indicating that they have remained unknown to the learning process.
- The position coincidence factor indicates the degree of positional coincidence between patient classification mask and disease classification masks in the vertical direction. The present concept is that parameters close to the disease process have higher coincidence factors than those depending on various specific genotypic and exposure conditions. Information on specific genotype and exposure conditions leading to disease (-> epidemiological data pattern analysis) may be obtained in this way.
- Data pattern analysis permits to evaluate the similarity of data patterns in a formalized way. Patients #137,139,141,142, 143,148 for example show a data pattern of: ++++ as opposed to: ++0++ for patients #138,144,146 in the above graph. The analysis of the data patterns may permit to detect typical genotypic and exposure conditions leading at various combinations finally to similar disease processes.
|© 2023 G.Valet|