Personal Verification/Identification via Analysis of the Peripheral ECG Leads: Influence of the Personal Health Status on the Accuracy

Table 5

Comparison between verification/identification accuracy achieved by the proposed method over the test dataset and the results reported by other authors with different databases used (db). The number of ECG recordings per patient (1 rpp for one and mrpp for more) and the acquisition interval (acq_int) on the same patient are reported.

Method	Database	Accuracy
Agrafioti and Hatzinakos, 2009 [5]	MIT-BIH normal sinus; MIT-BIH arrhythmia 1 rpp; PTB db, 13 healthy subjects mrpp	AccID = 96.2%; Sp_ver = 99%; Se_ver = 87%
Israel et al., 2005 [3]	Own db: 29 subjects; close in time recordings	AccID = 100%
Lourenço et al., 2011 [13]	Own db: 16 subjects; close in time recordings	AccID = 94.3%; Se = Sp = 87%
Sidek et al., 2012 [11]	Own db: 30 healthy subj., close in time recordings	AccID = 96.1%
Wang et al., 2008 [4]	MIT-BIH normal sinus 1 rpp; PTB db, 13 healthy subj. mrpp	AccID = 100%
Zhao et al., 2013 [1]	MIT-BIH ST change db, long-term ST db 1 rpp; PTB db, 12 healthy subj. mrpp	AccID (tot) = 95.6%; AccID (PTB) = 96%
Zokaee and Faez, 2012 [10]	MIT-BIH db 1 rpp; Own Holter, 50 subjects 1 rpp	AccID = 100%; AccID = 89%
Poree et al., 2011 [7]	Own db: 11 subjects, mrpp, acq_int = 16 months	AccID = 91.4%
Lee et al., 2012 [23]	Own db: 10 subjects, ~100 rpp within 3-month period	AccID = 99.5%
Wübbeler et al., 2007 [8]	db from 74 subjects, mrpp, acq_int = 16 months	AccID = 98.1%; Se = Sp = 97.2%
Our method; (based on assessment of ( + )/2)	Test ILSA db, 49 healthy subjects mrpp	AccID = 77.6%; Se_ver = 89.8%; Sp_ver = 83.9%
	Test PTB db, 14 healthy subjects mrpp	AccID = 92.9%; Se_ver = 100%; Sp_ver = 81.9%