Inter-Patient Atrial Flutter Classification Using FFT-Based Features and a Low-Variance Stacking Classifier

Stacked machine learning models to classify atrial disorders based on clinical ECG features: a method to predict early atrial fibrillation

OBJECTIVES

Atrial Tachycardia (AT) and Left Atrial Enlargement (LAE) are atrial diseases that are significant precursors to Atrial Fibrillation (AF). There are ML models for ECG classification; clinical features-based classification is required. The suggested work aims to create stacked ML models that categorize Sinus Rhythm (SR), Sinus Tachycardia (ST), AT, and LAE signals based on clinical parameters for AF prognosis.

METHODS

The classification was based on thirteen clinical parameters, such as amplitude, time domain ECG aspects, and P-Wave Indices (PWI), such as the ratio of P-wave length and amplitude ((P (ms)/P (µV)), P-wave area (µV*ms), and P-wave terminal force (PTFV1(µV*ms). Apart from classifying the ECG signals, the stacked ML models prioritized the clinical features using a pie formula-based technique.

RESULTS

The Stack 1 model achieves 99% accuracy, sensitivity, precision, and F1 score, while the Stack 2 model achieves 91%, 91%, 94%, and 92% for identifying SR, ST, LAE, and AT, respectively. Both stack models obtained a computational time of 0.06 seconds. PTFV1 (µV*ms), P (ms)/P (µV)), and P-wave area (µV*ms) were ranked as crucial clinical features.

CONCLUSION

Clinical feature-based stacking ML models may help doctors obtain insight into important clinical ECG aspects for early AF prediction.

Human Lower Limb Motion Capture and Recognition Based on Smartphones

Human motion recognition based on wearable devices plays a vital role in pervasive computing. Smartphones have built-in motion sensors that measure the motion of the device with high precision. In this paper, we propose a human lower limb motion capture and recognition approach based on a Smartphone. We design a motion logger to record five categories of limb activities (standing up, sitting down, walking, going upstairs, and going downstairs) using two motion sensors (tri-axial accelerometer, tri-axial gyroscope). We extract the motion features and select a subset of features as a feature vector from the frequency domain of the sensing data using Fast Fourier Transform (FFT). We classify and predict human lower limb motion using three supervised learning algorithms: Naïve Bayes (NB), K-Nearest Neighbor (KNN), and Artificial Neural Networks (ANNs). We use 670 lower limb motion samples to train and verify these classifiers using the 10-folder cross-validation technique. Finally, we design and implement a live detection system to validate our motion detection approach. The experimental results show that our low-cost approach can recognize human lower limb activities with acceptable accuracy. On average, the recognition rate of NB, KNN, and ANNs are 97.01%, 96.12%, and 98.21%, respectively.