NeoListen: detect neonatal crying in a hospital environment

The goal of the project is to create a machine learning algorithm to automatically detect the crying of babies in a challenging hospital environment. Because of a change in the hospital’s vision on how to treat neonates, they changed their policy from treating neonates in an open bay ward to single rooms. The AI envisions to notify the nurses of the Neonatal Intensive Care Unit (NICU) in case of a crying baby to be able to take action when they are not in the room.

Livia Popper
Jheronimus Academy of Data Science
Master in Data Science and Entrepreneurship

Over het initiatief / About the initiative

NeoListen: detect neonatal crying in a hospital environment

In welke fase zit jouw initiatief? / In what stage is your initiative?


Heb je jouw initiatief al gevalideerd? / Did you validate your initiative?

A proof of concept was done in January 2022. Then, a minimum viable product was developed in March 2022. Full implementation is currently in execution at Radboud UMC.

Meer informatie over jouw initiatief / More info about your initiative

A new single-patient room neonatology department opened at Radboud UMC in May 2022. In order to hear a baby cry, the only option for nurses is to leave the doors open in the new NICU. To hear a baby cry, there was a need for a crying detection solution that could deal with detecting cries in the midst of breathing machines, alarm sounds noise, amounting to a 50 decibels (dB) noisy environment.

The solution that we designed incorporates a full end-to-end system that continuously captures audio signal from the incubator where a preterm baby is placed. The signal is preprocessed to make it fit for the AI. The audio fragments are then used as input for the machine learning algorithm, which produces an output: 0, if baby is not crying, and 1, if baby is crying. In case of a baby crying, the system then sends a notification to the hospital phone of a nurse.

Wat zijn jouw volgende stappen om het verder te ontwikkelen? / What are your next steps to develop the initiative?

The next steps include the continuous improvement of the end-to-end solution by making use of feedback from nurses and real-world data, as we want to reuse the solution and to extend the original use-case to not only detect neonatal cry, but also to classify the cry of an infant as being part of one of the following categories: pain, hunger, or tiredness.

Wat heb je nodig om (nog meer) impact te maken met dit initiatief? / What do you need to make (more) impact with this initiative?

Financial support and brand awareness. To further develop the AI algorithm, more investments are needed. The project is also of great interest for multiple hospitals in the Netherlands.

The project is aiming to create a global impact, as it supports nurses in taking responsibility in challenging environments. Additionally, it helps nurses to work more efficiently, specifically in hospitals with nurse scarcity. Therefore, to accomplish that, attention for the innovation and brand awareness is needed to bring this project forward.