1 1

 0.56 0.5



 -0.5 -0.5

 -1 -1
 -1 -05 0 05 1 -1 -05 0 0.5 1



 40 40

 30 30

 20 20

 10 10

 0 0
 0 10 20 30 40 50 0 10 20 30 40 60
 (I) (ll)


Figure 3-6 Returns maps of the data for r = 2: (a) Flow; (b) Pressure; (I) Original data;
 (II) Test Set. For the test set, spontaneous breath is in red, mandatory in blue.

 As can be seen in Figure 3-5, the shapes of the trajectories for the chosen samples

replicate quite well the shapes of the whole data set (with slight shifts), and one can

immediately see the difference between the two breathing patterns, but also the

similarities: for the flow, signal on which is the segmentation is performed, the first and

third quarter of the map are very close, and the fourth quarter is a complete overlap. This

means that on the last part at least, the experts might not be able to discriminate between

the two modes, because they are not separable: it is the part where the exhaling takes

place, and since it is a passive mode for the patient and the ventilator, it follows the same

model for the two types of breath.

 It is also important to notice in Figure 3-3 that the data is a little noisy, and that

noise induces variations around the return map that can cause the instantaneous error of a