Page 199 - AI for Good-Innovate for Impact Final Report 2024

AI for Good-Innovate for Impact

Use case – 46: Improving early detection of neonatal asphyxia with

smartphone-based AI technologies 46 - BOU

Country: Bangladesh
Organization: Bangladesh Open University

Contact person: Samrat Kumar Dey (samrat.sst@ bou .ac .bd, +8801823267937)

46�1� Use case summary table

Domain healthcare
The problem to be addressed Development of more robust and reliable AI algorithms
for neonatal asphyxia detection on smartphones

Key aspects of the solution Voice sample analysis to identify asphyxia
Technology keywords Healthcare, asphyxia, voice, machine learning, deep
learning
Data availability Private right now.

Metadata (type of data) Audio sample from Mexico.
1049 audio samples.
1s samples.
Model Training and fine-tuning At the moment, using ML (regression models, ensemble
models)
May use DL in future (RNN-LSTM type models)
Testbeds or pilot deployments In 4 different hospitals from 4 different divisions.

46�2 Use-case description

46�2�1 Description

This use case explores leveraging smartphones and artificial intelligence (AI) for early detection
of neonatal asphyxia, a critical condition where newborns experience oxygen deprivation
during birth. Timely diagnosis is paramount for effective interventions and improved outcomes.
However, current methods in resource-constrained settings often rely on subjective assessments
or expensive equipment.

Intended Use: This AI-powered solution targets healthcare professionals, particularly midwives
and frontline workers, in low- and middle-income countries, where access to specialized

183