Leveraging Shape and Spatial Information for Spontaneous Preterm Birth Prediction

Paraskevas Pegios*, Emilie Pi Fogtmann Sejer, Manxi Lin, Zahra Bashir, Morten Bo Søndergaard Svendsen, Mads Nielsen, Eike Petersen, Anders Nymark Christensen, Martin Tolsgaard, Aasa Feragen

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

62 Downloads (Pure)

Abstract

Spontaneous preterm birth prediction from transvaginal ultrasound images is a challenging task of profound interest in gynecological obstetrics. Existing works are often validated on small datasets and may lack validation of model calibration and interpretation. In this paper, we present a comprehensive study of methods for predicting preterm birth from transvaginal ultrasound using a large clinical dataset. We propose a shape- and spatially-aware network that leverages segmentation predictions and pixel spacing information as additional input to enhance predictions. Our model demonstrates competitive performance on our benchmark, providing additional interpretation and achieving the highest performance across both clinical and machine learning baselines. Through our evaluation, we provide additional insights which we hope may lead to more accurate predictions of preterm births going forwards.

Original languageEnglish
Title of host publicationProceedings of The 4th International Workshop of Advances in Simplifying Medical Ultrasound (ASMUS)
Volume14337
PublisherSpringer Science and Business Media Deutschland GmbH
Publication date2023
Pages57-67
ISBN (Print)978-3-031-44520-0
ISBN (Electronic)978-3-031-44521-7
DOIs
Publication statusPublished - 2023
Event4th International Workshop of Advances in Simplifying Medical Ultrasound - Vancouver, Canada
Duration: 8 Oct 20238 Oct 2023

Workshop

Workshop4th International Workshop of Advances in Simplifying Medical Ultrasound
Country/TerritoryCanada
CityVancouver
Period08/10/202308/10/2023

Keywords

  • Spontaneous Preterm Birth
  • Transparency
  • Transvaginal Ultrasound

Fingerprint

Dive into the research topics of 'Leveraging Shape and Spatial Information for Spontaneous Preterm Birth Prediction'. Together they form a unique fingerprint.

Cite this