This example shows how the visualisation technique together with post cluster classification can be used to compare clinical groups. The figures below show the results of comparing the ambulatory manometry between patients with confirmed Barrett's oesophagus (red) and healthy controls (blue). Three simple post cluster classifications were used i.e. the overall plot shows Primary, Secondary and Non peristaltic events between the two groups.
It can easily be seen that the groups form different clusters, with a
small degree of overlap. It can be hypothesised that the outliers in the
healthy controls group (volunteers with no known problem) should be
screened or monitored for Barrett's i.e. this technique could be used a tool
to identify candidates at a risk of developing Barrett's that require
entry into a screening programme etc. The statistical tests
are shown below.
Mean number of primary swallows Barrett's = 334.6875, Controls = 1054.1
Results of Student T-test (= var) P = 2.5578e-004 Results of Student T-test (# var) P = 7.2077e-005
Mean number of secondary persistaltic Barrett's = 190.2500, Controls = 591.6923
Results of Student T-test (= var), P = 2.5578e-004 Results of Student T-test (# var), P = 7.2077e-005
Mean number of non peristaltic Barrett's = 3.5285e+003 Controls = 2.5925e+003
Results of Student T-test (= var) P = 2.5578e-004 Results of Student T-test (# var) P = 7.2077e-005
Primary = 1062, Secondary = 606, Unclassified = 2983
(distance between cluster centre and example)
d1 = 3.0402e+003 d2 = 1.9865e+003
Conclusion: Closer to normal than Barrett's
Back to http://www.medeng.net