Effects of distance transform choice in training with boundary loss

Breznik, EvaUppsala universitet,Bildanalys och människa-datorinteraktion,Avdelningen Vi3 (author)

Convolutional neural networks are the method of choice for many medical imaging tasks, in particular segmentation. Recently, efforts have been made to include distance measures in the network training, as for example the introduction of boundary loss, calculated via a signed distance transform. Using boundary loss for segmentation can alleviate issues with imbalance and irregular shapes, leading to a better segmentation boundary. It is originally based on the Euclidean distance transform. In this paper we investigate the effects of employing various definitions of distance when using the boundary loss for medical image segmentation. Our results show a promising behaviour in training with non-Euclidean distances, and suggest a possible new use of the boundary loss in segmentation problems.

Strand, Robin,1978-Uppsala universitet,Bildanalys och människa-datorinteraktion,Avdelningen Vi3(Swepub:uu)rostr275 (author)
Uppsala universitetBildanalys och människa-datorinteraktion (creator_code:org_t)

About the subject

ENGINEERING AND TECHNOLOGY: ENGINEERING AND ...; and Medical Engineer ...; and Medical Image Pr ...

Kungliga biblioteket hanterar dina personuppgifter i enlighet med EU:s dataskyddsförordning (2018), GDPR. Läs mer om hur det funkar här.
Så här hanterar KB dina uppgifter vid användning av denna tjänst.