Header

UZH-Logo

Maintenance Infos

Efficient Multi-class Fetal Brain Segmentation in High Resolution MRI Reconstructions with Noisy Labels


Payette, Kelly; Kottke, Raimund; Jakab, András (2020). Efficient Multi-class Fetal Brain Segmentation in High Resolution MRI Reconstructions with Noisy Labels. In: Hu, Y. Medical Ultrasound, and Preterm, Perinatal and Paediatric Image Analysis. ASMUS 2020, PIPPI 2020. Cham: Springer, 295-304.

Abstract

Segmentation of the developing fetal brain is an important step in quantitative analyses. However, manual segmentation is a very time-consuming task which is prone to error and must be completed by highly specialized individuals. Super-resolution reconstruction of fetal MRI has become standard for processing such data as it improves image quality and resolution. However, different pipelines result in slightly different outputs, further complicating the generalization of segmentation methods aiming to segment super-resolution data. Therefore, we propose using transfer learning with noisy multi-class labels to automatically segment high resolution fetal brain MRIs using a single set of segmentations created with one reconstruction method and tested for generalizability across other reconstruction methods. Our results show that the network can automatically segment fetal brain reconstructions into 7 different tissue types, regardless of reconstruction method used. Transfer learning offers some advantages when compared to training without pre-initialized weights, but the network trained on clean labels had more accurate segmentations overall. No additional manual segmentations were required. Therefore, the proposed network has the potential to eliminate the need for manual segmentations needed in quantitative analyses of the fetal brain independent of reconstruction method used, offering an unbiased way to quantify normal and pathological neurodevelopment.

Abstract

Segmentation of the developing fetal brain is an important step in quantitative analyses. However, manual segmentation is a very time-consuming task which is prone to error and must be completed by highly specialized individuals. Super-resolution reconstruction of fetal MRI has become standard for processing such data as it improves image quality and resolution. However, different pipelines result in slightly different outputs, further complicating the generalization of segmentation methods aiming to segment super-resolution data. Therefore, we propose using transfer learning with noisy multi-class labels to automatically segment high resolution fetal brain MRIs using a single set of segmentations created with one reconstruction method and tested for generalizability across other reconstruction methods. Our results show that the network can automatically segment fetal brain reconstructions into 7 different tissue types, regardless of reconstruction method used. Transfer learning offers some advantages when compared to training without pre-initialized weights, but the network trained on clean labels had more accurate segmentations overall. No additional manual segmentations were required. Therefore, the proposed network has the potential to eliminate the need for manual segmentations needed in quantitative analyses of the fetal brain independent of reconstruction method used, offering an unbiased way to quantify normal and pathological neurodevelopment.

Statistics

Citations

Dimensions.ai Metrics
9 citations in Web of Science®
8 citations in Scopus®
Google Scholar™

Altmetrics

Additional indexing

Item Type:Book Section, refereed, original work
Communities & Collections:04 Faculty of Medicine > University Children's Hospital Zurich > Medical Clinic
04 Faculty of Medicine > University Children's Hospital Zurich > Clinic for Surgery
Dewey Decimal Classification:610 Medicine & health
Scopus Subject Areas:Physical Sciences > Theoretical Computer Science
Physical Sciences > General Computer Science
Language:English
Date:1 January 2020
Deposited On:15 Dec 2021 14:37
Last Modified:26 Jun 2024 01:46
Publisher:Springer
Series Name:Lecture Notes in Computer Science
Number:1234
ISSN:0302-9743
ISBN:9783030603335
OA Status:Closed
Publisher DOI:https://doi.org/10.1007/978-3-030-60334-2_29
Related Items:
Full text not available from this repository.