Detection Quality Indices for Improved Heart Beat Assessment in Non-Invasive Fetal ECG

In this work, we establish the novel concept of detection quality indices (DQI) for the assessment of non-invasive fetal heartbeat detections. We grouped the DQIs in four categories concerning the detection properties they address and used these DQIs as features to classify correct detections and two detection errors with a decision tree. Our classifier was able to distinguish between correct detection and two detection errors (f1 = 0.90). By excluding the detection errors, we were able to improve the error of time-domain heart rate variability parameters and approximated entropy compared to manual detection by 35 percentage points on average in a first clinical test setting.