Multi-scale module-based algorithm for skin cancer image segmentation

For the problems of intra-class variability, inter-class similarity, and unbalanced dataset of skin cancer images, this paper proposes a segmentation method based on attention fusion network. The segmentation network is based on U-Net network as a segmentation network, and the segmentation accuracy...

Full description

Saved in:
Bibliographic Details
Published inIEEE ... Information Technology and Mechatronics Engineering Conference (ITOEC ... ) (Online) Vol. 7; pp. 320 - 323
Main Authors Shen, Tongping, Menchita, Dumlao
Format Conference Proceeding
LanguageEnglish
Published IEEE 15.09.2023
Subjects
Attention Mechanism
Decoding
Feature extraction
Fuses
Image segmentation
Interference
Mechatronics
Segmentation
Semantics
skin cancer
U-Net
Online AccessGet full text
ISSN2693-289X
DOI10.1109/ITOEC57671.2023.10291203

Cover

More Information
Summary:For the problems of intra-class variability, inter-class similarity, and unbalanced dataset of skin cancer images, this paper proposes a segmentation method based on attention fusion network. The segmentation network is based on U-Net network as a segmentation network, and the segmentation accuracy is enhanced by multi-scale modules that fuse features at different levels during feature extraction and give certain weights to important target features, thus strengthening the importance of channel and spatial pixel features in the decoder. The proposed model was trained and tested on the ISIC2018 dataset, and the experimental results achieved an overall excellent result with Precision reaching 0.9, Recall reaching 0.9, and F1 reaching 92.1%.
ISSN:2693-289X
DOI:10.1109/ITOEC57671.2023.10291203