Table 7 Comparison of noise detection performance between the proposed and existing methods.
From: Design and evaluation of a knowledge-based ECG noise filtering framework
Methods and Year | Type | Data | No. of datasets | Acc. (%) | Sen. (%) | Sp. (%) | F1 (%) |
|---|---|---|---|---|---|---|---|
CEEMD (fixed threshold) | CINC-2011, MIT-BIH arrhythmia ,MIT-STC, fantasia, PTB | 5 | 99.35 | 99.5 | 98.88 | N/R | |
ML | Private dataset | 1 | 95.63 | 97.41 | 86.50 | N/R | |
ML | CINC-2011, CINC-2014 , TeleECG, Private dataset | 4 | 97.15 | 89.08 * | 88.15 * | N/R | |
DL | CINC-2011 | 1 | 93.09 | 97.67 | 84.73 | 84.72 | |
ML | Private dataset | 1 | 95.23 * | N/R | N/R | N/R | |
Wavelet (fixed threshold) | CINC-2011 | 1 | 85.75 | 92.00 * | 67.00 * | N/R | |
DL Two data input (Raw and feature) | MIT-BIH noise stress test, MIT-BIH arrhythmia | 2 | 100.00 | 99.00 | 100.00 | N/R | |
DL | Private, CINC-2011 | 2 | 94.55 | 85.68 | 97.02 | 94.36 | |
DL | TeleECG,MIT-BIH arrhythmia, , BUT QDB, NSRDB | 4 | 97.00 | 95.50 | 97.00 | N/R | |
Proposed | ML | ECG-ID, BIDMC, CINC-2011, CINC-2014, MIT/BIH noise stress, TeleECG, MIT-BIH arrhythmia | 7 | 99.82 | 99.27 | 99.94 | 99.88 |
