Work

Infant Hippocampal Segmentation in Ultra-Low-Field MRI Using External Datasets with Diverse Field Strengths

Year: 2025

Type: conference-paper

Abstract: Ultra-low-field (uLF) MRI is a promising technology for expanding access to MRI in resource-limited settings, but it presents significant image analysis challenges, particularly in the segmentation of small brain structures. One such structure is the hippocampus, which is crucial for monitoring the neurodevelopment after birth. In this study, we developed a hippocampal segmentation method using the nnUNet framework, exploring various training strategies, including using additional labeled high-field (HF) and unlabeled low-field (LF) data. Our results show that integrating external datasets improves segmentation accuracy over using uLF data alone, even with substantial differences in imaging parameters and field strengths. This approach highlights the importance of leveraging diverse datasets to enhance the performance of segmentation models in low-quality imaging modalities, potentially leading to better diagnostic capabilities in challenging clinical environments.

Source: Lecture notes in computer science

DOI: 10.1007/978-3-031-83008-2_3

Authors: Weichen Zhou, Jingyu Li, Xi Wang, Yuwan Wang, Mengye Lyu

Institutions: Shenzhen Technology University, Chinese University of Hong Kong, Zhejiang Lab

Language: English

Topic: Medical Image Segmentation Techniques

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Good health and well-being

Open Access status: hybrid

Estimated APC paid: $5,392