Osaka Metropolitan University was the only Japanese institution among 30 participating institutions across 20 countries and territories


Participating regions in the development of ai for neonatal patients using adult medical AI

Title: Participating regions in the development of ai for neonatal patients using adult medical AI

Caption: Osaka Metropolitan University, the only Japanese institution that participated in the study, contributed by collecting and annotating chest X-ray images from their medical school.

Credit: Osaka Metropolitan University



Dr. Yasuhito Mitsuyama and Professor Daiju Ueda of the Graduate School of Medicine at Osaka Metropolitan University (OMU) participated in an international collaborative study into applying adult medical AI to neonatal intensive care units (NICUs). The study was led by Harvard Medical School and involved 30 institutions from around the world. OMU was the only participating institution from Japan.

Rapid and accurate image diagnosis in NICUs is a critical issue that can directly affect newborns’ lives. However, the limited availability of neonatal medical imaging data has made it difficult to develop highly accurate artificial intelligence (AI) systems.

To address this limitation, the study evaluated the effectiveness of a continual learning approach that applies an AI model originally developed using adult chest X-ray data to NICU patients.

By sequentially training the model with the small number of neonatal images available at each participating institution, the researchers were able to significantly improve the accuracy of detecting the positions of endotracheal tube tips and tracheal bifurcation.

These findings suggest that existing AI models can be adapted for neonatal care using a privacy-preserving approach that does not require the collection of large amounts of new data.

Researchers at OMU contributed to the study by collecting and annotating chest X-ray images from their medical school, supporting the evaluation of the model's performance.

The findings were published in Nature Biomedical Engineering.

Paper information

Journal: Nature Biomedical Engineering
Title: Repurposing adult AI radiograph models for neonatal care through continual learning: an international multicentre study
DOI: 10.1038/s41551-026-01812-6
Author(s): Vishwanatha M. Rao, et al.
Publication date: 13 October 2026
URL: https://doi.org/10.1038/s41551-026-01812-6

Contact

Yasuhito Mitsuyama
Graduate School of Medicine
Email: g21324t[at]omu.ac.jp

*Please change [at] to @.