Real-time quality feedback on Doppler data for community midwives using edge-AI

Area of Science:

• Medical Technology
• Artificial Intelligence
• Global Health

Background:

• Fetal Doppler monitoring is crucial for maternal and fetal health, especially in low-resource settings.
• Accurate fetal Doppler data is essential for timely diagnosis of pregnancy-related conditions.
• Existing data collection methods can be limited by signal quality and lack of real-time feedback.

Purpose of the Study:

• To develop and validate a technical framework for real-time fetal Doppler data quality assessment using deep learning and edge-AI.
• To integrate this framework into a low-cost, edge-computing system for use in rural, low-resource environments.
• To improve the accuracy and reliability of fetal cardiac signal acquisition for clinical decision support.

Main Methods:

• A deep neural network was trained on segmented fetal Doppler recordings (3.75s intervals) categorized into five quality levels.
• The model was trained and validated using a dataset from rural Guatemala and tested on a dataset from Leipzig, Germany.
• The algorithm was implemented within an Android mHealth application for real-time feedback during signal acquisition.

Main Results:

• The deep neural network achieved high performance, with a micro F1-score of 97.4% and macro F1-score of 94.2% on the Guatemala dataset.
• The model demonstrated generalization capabilities, achieving a 93.3% F1-score for 'Good' quality segments on the Leipzig dataset.
• Real-time feedback during signal acquisition significantly improved data quality at the source.

Conclusions:

• The developed edge mHealth solution provides a scalable and effective method for real-time fetal Doppler data quality assessment.
• This technology has the potential to significantly enhance maternal and fetal health monitoring in the Global South.
• The integration of mobile technology, AI, and healthcare offers a promising approach to addressing global health challenges.