Table 4 Heartbeat classification methods. ISCH – ischemic heartbeats; PTCA - percutaneous transluminal coronary angioplasty; Se, Se, Acc – sensitivity, specificity and accuracy, respectively.
Type of heartbeats | Features | Classifier | Classification performance | Authors |
|---|---|---|---|---|
NOR, particular minutes of ISCH | Morphological (QRS loop, ST-change vector magnitude, spatial ventricular gradient features) | Linear DFA | Se = 95.4% | Correa et al.3 |
Sp = 95.2 (for the 4th min of the PTCA) | ||||
NOR, mild ISCH | High-frequency QRS | Logistic regression | Se = 46%, Sp = 87% | Sharir et al.37 |
NOR, moderate/severe ISCH | Se = 69%, Sp = 86% | |||
NOR, mild ISCH | Morphological (Q and R amplitudes, QRS slopes and energy, repolarization duration, normalized ST, etc.) | Logistic regression | Se = 77%, Sp = 88% | Llamedo et al. 2013 cinc38 |
NOR, moderate/severe ISCH | Se = 76%, Sp = 90% | |||
NOR, ISCH | Morphological (QRS slopes) | Linear DFA | Se = 24%, Sp = 93% | Firoozabadi et al.4 |
NOR, ISCH | Morphological (samples of heartbeats) | k-NN, SVM | Acc(k-NN) = 94.6% | Murthy et al.5 |
Acc(linear SVM)=91.5% | ||||
Acc(RBF SVM)=95.3% | ||||
NOR, ISCH | Morphological | SVM | Se = 94.8%, Sp = 99.5% | Tseng et al.6 |
NOR, VPB and 2 others | Morphological, RR intervals, higher-order statistics | Linear DFA, | DFA: Acc(NOR)=88.6% | Doquire et al.10 |
SVM | Acc(VPB)=80.6% | |||
SVM: Acc(NOR)=75.9% | ||||
Acc(VPB)=85.1% | ||||
NOR, VPB and 1 other | AR model coefficients and non-linear features | Linear DFA | Acc up to 88% | Balli et al.9 |
NOR, VPB | Morphological | SVM | Acc = 88.9% | Alajlan et al.7 |
NOR,VPB + 3 others (except ISCH) | Higher order statistics of wavelet coefficients | k-NN | Se(NOR) = 97.5% | Kutlu et al.12 |
Sp(NOR) = 95.3% | ||||
Se(VPB) = 87.6% | ||||
Sp(VPB) = 96.8% | ||||
NOR, VPB + 5 others (except ISCH) | Spectral | k-NN | Acc = 97.0% | Arif et al.8 |
NOR,VPB + 3 others (except ISCH) | Higher order statistics | k-NN | Se(NOR) = 100% | Karimifard et al.11 |
Sp(NOR) = 99.9% | ||||
Se(VPB) = 94.1% | ||||
Sp(VPB) = 99.8% | ||||
NOR, VPB | Kalman filter, polarogram | NB | Acc = 98.8% | Sayadi et al.13 |
NOR, VPB+10 others (except ISCH) | Morphological, spectral and statistic | SVM | Acc = 98.9% | Shen et al.14 |
NOR, ISM, ISE, VPB | Morphological (AreaD) | k-NN | Acc = 98.6% | Proposed method |
NOR, ISM, ISE, VPB | Morphological (MorphD) | SVM | Acc = 93.5% | Proposed method |
