This repository hosts the scripts used for training and deploying the deep learning model for quality classification of 1-D Doppler Ultrasound signals. The model is detailed in the paper presented at the Machine Learning for Health (ML4H) 2023 conference: M. Motie-Shirazi, R. Sameni, P. Rohloff, N. Katebi, G. D. Clifford, "Point-of-Care Real-Time Signal Quality Assessment for Fetal Doppler Ultrasound Using a Deep Learning Approach", Machine Learning for Health, New Orleans, December 2023.
Please cite this work when using the work in this repository
- This Python script includes the deep learning model training procedure for signal quality classification of Doppler recordings.
- A Python script that provides auxiliary functions essential for the training of the model.
- A Jupyter Notebook that illustrates how to load and employ the model for classification tasks.
- Contains the trained deep learning model utilized for quality classification.
- Includes a 3.75-second signal sample from each of the quality classes: 'Good', 'Poor', 'Interference', 'Talking', and 'Silent', which can be used to test the model.