Real-time polyp segmentation using U-net with IoU loss
Colonoscopy is the third leading cause of cancer deaths worldwide. While automated segmentation methods can help detect polyps and consequently improve their surgical removal, the clinical usability of these methods requires a trade-off between accuracy and speed. In this work, we exploit the tradit...
| Main Authors: | , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
CEUR-WS.org Team
2021
|
| _version_ | 1826312517419270144 |
|---|---|
| author | Batchkala, G Ali, S |
| author_facet | Batchkala, G Ali, S |
| author_sort | Batchkala, G |
| collection | OXFORD |
| description | Colonoscopy is the third leading cause of cancer deaths worldwide. While automated segmentation methods can help detect polyps and consequently improve their surgical removal, the clinical usability of these methods requires a trade-off between accuracy and speed. In this work, we exploit the traditional U-Net methods and compare different segmentation-loss functions. Our results demonstrate that IoU loss results in an improved segmentation performance (nearly 3% improvement on Dice) for real-time polyp segmentation. |
| first_indexed | 2024-04-09T03:55:45Z |
| format | Conference item |
| id | oxford-uuid:13e3a211-4f1c-4782-a5ac-c84289374c78 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-04-09T03:55:45Z |
| publishDate | 2021 |
| publisher | CEUR-WS.org Team |
| record_format | dspace |
| spelling | oxford-uuid:13e3a211-4f1c-4782-a5ac-c84289374c782024-03-14T16:08:07ZReal-time polyp segmentation using U-net with IoU lossConference itemhttp://purl.org/coar/resource_type/c_5794uuid:13e3a211-4f1c-4782-a5ac-c84289374c78EnglishSymplectic ElementsCEUR-WS.org Team2021Batchkala, GAli, SColonoscopy is the third leading cause of cancer deaths worldwide. While automated segmentation methods can help detect polyps and consequently improve their surgical removal, the clinical usability of these methods requires a trade-off between accuracy and speed. In this work, we exploit the traditional U-Net methods and compare different segmentation-loss functions. Our results demonstrate that IoU loss results in an improved segmentation performance (nearly 3% improvement on Dice) for real-time polyp segmentation. |
| spellingShingle | Batchkala, G Ali, S Real-time polyp segmentation using U-net with IoU loss |
| title | Real-time polyp segmentation using U-net with IoU loss |
| title_full | Real-time polyp segmentation using U-net with IoU loss |
| title_fullStr | Real-time polyp segmentation using U-net with IoU loss |
| title_full_unstemmed | Real-time polyp segmentation using U-net with IoU loss |
| title_short | Real-time polyp segmentation using U-net with IoU loss |
| title_sort | real time polyp segmentation using u net with iou loss |
| work_keys_str_mv | AT batchkalag realtimepolypsegmentationusingunetwithiouloss AT alis realtimepolypsegmentationusingunetwithiouloss |