Learning to segment from object thickness annotations
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F23%3A00366106" target="_blank" >RIV/68407700:21230/23:00366106 - isvavai.cz</a>
Result on the web
<a href="https://doi.org/10.1109/ISBI53787.2023.10230621" target="_blank" >https://doi.org/10.1109/ISBI53787.2023.10230621</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1109/ISBI53787.2023.10230621" target="_blank" >10.1109/ISBI53787.2023.10230621</a>
Alternative languages
Result language
angličtina
Original language name
Learning to segment from object thickness annotations
Original language description
Measuring object size is fast and a standard part of many radiological evaluation procedures. We describe a deep learning segmentation method that can be trained on a small number of pixel-wise reference segmentation and then fine-tuned from the weak annotations of the object thickness. The difficulty is in the non-differentiability of the thickness function defined using the pixel-wise distance transform. We overcome it by optimizing the expected value of the loss function after the injection of a virtual random noise. Further speedup is possible using the properties of the distance transform. We demonstrate the benefit of the proposed method on ultrasound images of the carotid artery. The fine-tuning improves the performance by about 10% IoU.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
Result was created during the realization of more than one project. More information in the Projects tab.
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)
Others
Publication year
2023
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI)
ISBN
978-1-6654-7358-3
ISSN
1945-7928
e-ISSN
1945-8452
Number of pages
4
Pages from-to
—
Publisher name
IEEE Signal Processing Society
Place of publication
New Jersey
Event location
Cartagena de Indias
Event date
Apr 18, 2023
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
001062050500298