MIT-BIH Polysomnographic Database

There are many other physiological bases in PhysioBank. We are going to apply the knowledge of the tools of PhysioToolkit to the MIT-BIH Polysomnographic Database. It is a base including sleep recordings. We copy the files whose base-name is `slp01a'. We begin by checking that the files are in the directory:

-->chdir('/home/j/physionet')
 ans  =
 
    0.  
 
-->unix_g('ls slp01a*.*')
 ans  =
 
!slp01a.dat  !
!            !
!slp01a.ecg  !
!            !
!slp01a.hea  !
!            !
!slp01a.st   !
!            !
!slp01a.xws  !

If everything functions properly we should be able to read `slp01a.hea' with the command:

-->unix_x('cat slp01a.hea')

And the result is

slp01a 4 250/0.033333333(94) 1800000 23:07:00 19/1/1989
slp01a.dat 212 -200/mV 12 0 -17 59911 0 ECG
slp01a.dat 212 4.77778(-477)/mmHg 12 0 -248 19332 0 BP
slp01a.dat 212 -6430/mV 12 0 252 49594 0 EEG (C4-A1)
slp01a.dat 212 690/l 12 0 -180 912 0 Resp (sum)
# 44 M 89 32-01-89

There is a lot of information here. Among other things, we can deduce that it is the recording of a 44-year-old male and includes four signals: ECG, BP, EEG and RESP. It has been sampled at 250 Hz and includes 1,800,000 samples per channel.

We can try to recover the data with the usual procedure

-->unix ('rdsamp -r slp01a -p > dummy');

-->stacksize(20000000)
 
-->x = fscanfMat('dummy');
 
-->wind  = 1:15000;
 
-->for k = 1:4; xsetech([0,(k-1)/4,1,1/4]);plot2d(x(wind,1),x(wind,1+k));end

And the result is shown in the figure  :

Figure: Plotting of the first 60 seconds of slp01a.dat

We are also interested in the other files.

-->unix_x('rdann -r slp01a -a ecg')

produces

[23:07:00.076 19/01/1989]       19     N    0    0    0
[23:07:00.936 19/01/1989]      234     N    0    0    0
[23:07:01.828 19/01/1989]      457     N    0    0    0
[23:07:02.748 19/01/1989]      687     N    0    0    0
[23:07:03.644 19/01/1989]      911     N    0    0    0
[23:07:04.548 19/01/1989]     1137     N    0    0    0
[23:07:05.440 19/01/1989]     1360  ...

It is the usual file containing the location and type of the beats. And the other one

 
-->unix_x('rdann -r slp01a -a st')

contains information about sleep apnea

[23:07:00.004 19/01/1989]        1     "    0    0    0	4 LA LA
[23:07:30.000 19/01/1989]     7500     "    0    0    0	4 LA
[23:08:00.000 19/01/1989]    15000     "    0    0    0	4 LA
[23:08:30.000 19/01/1989]    22500     "    0    0    0	4 L L L
[23:09:00.000 19/01/1989]    30000     "    0    0    0	4 L
[23:09:30.000 19/01/1989]    37500     "    0    0    0	4 L
[23:10:00.000 19/01/1989]    45000...

So, with relatively little effort we have been able to read signals from very different origin that can be stored as Scilab variables in such a way that all the power of Scilab detailed in the companion tutorial can be applied.