Deep Learning for Automated Wound Classification And Segmentation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917754570801152 |
|---|---|
| author | Jahangir, Md. Zihad Bin Akter, Sumaiya Nasim, MD Abdullah Al Gupta, Kishor Datta George, Roy |
| author_facet | Jahangir, Md. Zihad Bin Akter, Sumaiya Nasim, MD Abdullah Al Gupta, Kishor Datta George, Roy |
| contents | Wounds, such as foot ulcers, pressure ulcers, leg ulcers, and infected wounds, come up with substantial problems for healthcare professionals. Prompt and accurate segmentation is crucial for effective treatment. However, contemporary methods need an exhaustive model that is qualified for both classification and segmentation, especially lightweight ones. In this work, we tackle this issue by presenting a new architecture that incorporates U-Net, which is optimized for both wound classification and effective segmentation. We curated four extensive and diverse collections of wound images, utilizing the publicly available Medetec Dataset, and supplemented with additional data sourced from the Internet. Our model performed exceptionally well, with an F1 score of 0.929, a Dice score of 0.931 in segmentation, and an accuracy of 0.915 in classification, proving its effectiveness in both classification and segmentation work. This accomplishment highlights the potential of our approach to automating wound care management. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_11064 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Deep Learning for Automated Wound Classification And Segmentation Jahangir, Md. Zihad Bin Akter, Sumaiya Nasim, MD Abdullah Al Gupta, Kishor Datta George, Roy Image and Video Processing Wounds, such as foot ulcers, pressure ulcers, leg ulcers, and infected wounds, come up with substantial problems for healthcare professionals. Prompt and accurate segmentation is crucial for effective treatment. However, contemporary methods need an exhaustive model that is qualified for both classification and segmentation, especially lightweight ones. In this work, we tackle this issue by presenting a new architecture that incorporates U-Net, which is optimized for both wound classification and effective segmentation. We curated four extensive and diverse collections of wound images, utilizing the publicly available Medetec Dataset, and supplemented with additional data sourced from the Internet. Our model performed exceptionally well, with an F1 score of 0.929, a Dice score of 0.931 in segmentation, and an accuracy of 0.915 in classification, proving its effectiveness in both classification and segmentation work. This accomplishment highlights the potential of our approach to automating wound care management. |
| title | Deep Learning for Automated Wound Classification And Segmentation |
| topic | Image and Video Processing |
| url | https://arxiv.org/abs/2408.11064 |