The aim of this study was to develop a method that could automatically evaluate the quality of ECG recordings. In several situations, people performing the recording don't have the knowledge to evaluate the quality of the ECG and an immediate feedback about it would be useful. Since there is not a consensus on how to define and quantify ECG quality, we applied self learning techniques starting from a set (N= 1 OOO) of randomly selected ECGs from our internal repository. The full set of ECGs was blindly flagged by an expert cardiologist and subsequently analyzed by AMPS software which automatically computes a set of quality metrics. These quality parameters were used to train a neural network and build a decision tree. The performance of the proposed solutions were evaluated using the mean squared error (MSE) between expected results (from the ECGs set) and obtained results (from neural network and decision tree). The MSE resulting from the neural network and the decision tree were O.O1 and 0.004, respectively, indicating an error in range of 1%.
[1]
J. Ross Quinlan,et al.
C4.5: Programs for Machine Learning
,
1992
.
[2]
Alberto Maria Segre,et al.
Programs for Machine Learning
,
1994
.
[3]
Warren S. Sarle,et al.
Stopped Training and Other Remedies for Overfitting
,
1995
.
[4]
Lars Johannesen,et al.
Automatic ECG quality scoring methodology: mimicking human annotators
,
2012,
Physiological measurement.
[5]
Günter Schreier,et al.
QRS detection based ECG quality assessment
,
2012,
Physiological measurement.
[6]
Ian H. Witten,et al.
The WEKA data mining software: an update
,
2009,
SKDD.
[7]
T.,et al.
Training Feedforward Networks with the Marquardt Algorithm
,
2004
.
[8]
H Naseri,et al.
An expert electrocardiogram quality evaluation algorithm based on signal mobility factors
,
2013,
Journal of medical engineering & technology.
[9]
Mohammad Bagher Menhaj,et al.
Training feedforward networks with the Marquardt algorithm
,
1994,
IEEE Trans. Neural Networks.