ECG Compression - ECG Acquisition and Automated Remote Processing - page 110

Table 5.5 PRD and CR values for compression using c-db data
Patient file ID in
PhysioNet

Lead 1 Lead 2

CR PRD

(%)

PRDN
(%)

QS E max CR PRD

(%)

PRDN
(%)

QS E max

08730_03 26.97 1.63 1.75 16.54 0.009 25.83 1.66 1.671 15.56 0.014

08730_04 25.33 1.42 1.46 17.83 0.043 25.96 1.31 1.525 19.81 0.019

11950_03 41.12 3.48 7.99 11.81 0.076 39.21 11.98 12.87 3.27 0.117

11247_01 38.11 9.83 9.96 3.87 0.155 33.49 3.42 5.15 9.79 0.040

11247_03 40.16 7.98 8.21 5.03 0.205 40.27 9.73 9.812 4.13 0.030

Table 5.6 PRD and CR values for compression using mit-db data
Patient file ID in PhysioNet
with lead no.

CR PRD (%) PRDN (%) QS Maximum

error

101-v1 31.82 1.231 8.359 25.84 0.009

102-v2 32.14 0.435 0.445 73.88 0.043

104-v2 58.64 2.49 2.97 23.55 0.040

105-v1 37.90 2.15 3.228 17.62 0.008

Table 5.5 represents some of the test results with c-db data using this down-
sampling technique. An average CR, PRD, PRDN, and E max of 39.12, 4.54, 7.42,
and 0.07 are found, respectively.
In the next stage, mit-db data are also used for validating the modified algo-
rithm. Out of 25 individual leads tested, an average CR, PRD, PRDN, QS, and
E max obtained are 43.54, 1.73, 3.14, and 0.052, respectively. Table 5.6 shows some
of the test results with mit-db data.
Finally, ptb-db data are used for testing the algorithm performance. An average
CR and PRD values of 52.04 and 2.12, respectively, are found. Figure 5.8 rep-
resents one such c-db and mit-db record being decompressed and compared with
original sample.
DDF parameters are also evaluated for estimation of clinical signatures in the
reconstructed data. Table 5.7 provides an estimate of DDF parameters before
compression and after reconstruction, computed over 2 beats, for 75 mit-db and
c-db leads tested using down-sampled data used for compression.

5.3.5 Compression Performance with Thresholding on First-
Difference Array

To find the suitability of implementing the compression using 8-bit microcon-
troller, the PhysioNet data are quantized to 8-bit resolution. Since the PhysioNet
data are with 16-bit resolution, to enhance the CR and a threshold value is
included while computing the first difference array. Two threshold values 0.39%