Irregular Breathing Classification from Multiple Patient Datasets - Prediction and Classification of Respiratory Motion - page 117

we will change the discrimination threshold by the period of observation (T i )to
validate the performance of the proposed binary classifier system.
To predict the irregular breathing patterns from the patient datasets, we may
evaluate the classification performance by showing the following two ROC analysis:
As the first ROC, we may increase the threshold value n m defined in ( 6.18 )in
Sect. 6.3.4 , from 0.1 to 0.99. By changing the observation period T i of 900, 300,
and 100 s, the system may include the irregular breathing patterns extracted under
the different parameters of n m . Specifically, depending on the observation period
T i , we would like to adjust the threshold value n m for the ROC evaluation of the
proposed classifier.
As the second ROC, we may increase the regular threshold Wt ð Þ so that the
patient datasets with the ratio c ðÞ of patient i may be changed from true negative
to true positive. For the analysis based on the regular threshold, we extract the ratio
c ðÞ of patient i by changing the observation period T i of 900, 300, and 100 s. The
regular threshold Wth can be variable from 0.1 to 0.99, especially by changing the
regular threshold Wth of 0.80, 0.85, and 0.90, defined in Sect. 6.4.1 . Depending on
the regular threshold Wth ð Þ , ROC is analyzed for the performance of the proposed

classifier.

6.5 Experimental Results

6.5.1 Breathing Motion Data

Three channel breathing datasets with a sampling frequency of 26 Hz are used to
evaluate the performance of the proposed irregular breathing classifier. Here each
channel makes a record continuously in three dimensions for 448 patients datasets.
The breathing recording time for each patient is distributed from 18 min to 2.7 h,
with 80 min as the average time at the Georgetown University Cyberknife treat-
ment facility. In Fig. 6.5 we restricted the breathing recording times to discrete
values with the unit of five minutes. That means 18 min recording time is quan-
tized to 20 min for a variable quantity. Figure 6.5 shows the frequency distribution
of breathing recording time.
The minimum and the maximum recording times are 18 and 166 min in
Fig. 6.5 . To extract the feature extraction metrics in Table 6.1 , therefore, we
randomly selected 18 min samples from the whole recording time for each
breathing dataset because the minimum breathing recording time is 18 min. That
means we use 28,080 samples to get the feature extraction metrics for each
breathing dataset. Meanwhile, every dataset for each patient is analyzed to predict
the irregular breathing patterns. That means we inspect all the datasets to detect the
irregular pattern w ðÞ within the entire recording time. The detected irregular
patterns can be used to calculate the true positive/negative ranges R T i R TN

and

i

the ratio c ðÞ for the patients.