Fig 1.
Example for the application of this method for detecting AF.
(A) The original HR sequence hrn; (B) The distribution of symbols syn; (C) The relevant word sequence wvn of syn in (b), and (D) The distribution of SE (A).
Fig 2.
Overview of the beat-by-beat AF detection algorithm.
Fig 3.
Receiver operating characteristic for the present algorithm when the training set of LTAFDB database is applied with threshold values from 0.0 to 1.0 in increments of 0.001.
The calculated value for the area under the blue curve is 0.9845.
Table 1.
Summary of classification performance for three different testing databases with various cases (with the threshold of 0.639).
Fig 4.
Probability histogram for annotated AF and non-AF beats of AFDB set.
Table 2.
Overview of performance comparison of previous algorithm using the same databases.
Fig 5.
Receiver operating characteristic (ROC) curves for the present and our previous algorithms when the AFDB, AFDB†, AFDB‡ and NSRDB databases are tested with various situations.
(A) ROC curves of the AFDB set; (B) ROC curves of the AFDB† database († indicates records “00735” and “03665” excluded); (C) ROC curves of the AFDB‡ database (‡ indicates records “04936” and “05091” excluded); (D) Results of the AFDB+NSRDB database, (E) Results of the AFDB†+NSRDB database and (F) Results of the AFDB‡+NSRDB database. The calculated values for the area under the curves are listed in Table 3.
Table 3.
The areas under receiver operating characteristic (ROC) curves for the present and our previous algorithms when AFDB, AFDB†, AFDB‡ and NSRDB databases are tested with various situations.
Table 4.
The computation time of the processing of this method.