Skin Lesion Segmentation with Semantic SAM: Pros and Cons

Sevda GÜL, Bekir Murat AYDIN, Devrim AKGÜN, Rabia ÖZTAŞ KARA, Gökçen ÇETİNEL

Abstract


The Segment Anything Model (SAM) was introduced for the first time in April 2023 and gained popularity in a short time. One of the critical application areas of SAM is medical image segmentation. The performance of SAM in various medical image segmentation tasks is still being investigated. The main specifications of SAM can be summarized as follows: i) Excellent generalization on common scenes, ii) Strong prior knowledge requirement, iii) Less effective in low-contrast applications, iv) Limited understanding of professional data, v) Performance degradation in smaller and irregular objects.

In this study, our goal is to investigate the performance of SAM for skin lesion segmentation tasks. For this purpose, two comprehensive databases are constructed. The first database includes 3463 public skin lesion images, and the second database consists of 773 private skin lesion images taken from Sakarya University Training and Research Hospital with ethical permission. The performance evaluation compares manually determined skin lesion regions with automatic segmentation regions achieved by SAM through Intersection over Union (IoU) and Dice metrics.   In the presented study, a post-processing step is also applied to increase the skin lesion segmentation ability of SAM, and the results are compared with commonly used deep learning-based segmentation methods.


Keywords


Skin lesion segmentation; Segment Anything model; Deep learning

Full Text:

PDF

References


Khattar, S., Kaur, R., & Gupta, G. (2023). A Review on Preprocessing, Segmentation and Classification Techniques for Detection of Skin Cancer. 2nd Edition of IEEE Delhi Section Owned Conference, DELCON 2023- Proceedings. https://doi.org/10.1109/DELCON57910.2023.10127546

World Health Organization (WHO)/International Agency, “Cancer Today”, 27.10.2023, Available: https://gco.iarc.fr/today/online-analysis

American Cancer Society, Skin Cancer, 27.10.2023, Available: https://www.cancer.org/cancer/types/skin-cancer.html

Mirikharaji, Z., Abhishek, K., Bissoto, A., Barata, C., Avila, S., Valle, E., Celebi, M. E., & Hamarneh, G. (2023). A Survey on Deep Learning for Skin Lesion Segmentation. https://doi.org/10.1016/j.media.2023.102863

C. Akyel and N. Arıcı, “LinkNet-B7: Noise Removal and Lesion Segmentation in Images of Skin Cancer,” Mathematics, vol. 10, no. 5, pp. 736, Feb. 2022, doi: https://doi.org/10.3390/math10050736

L. Li and W. Seo, "Deep Learning and Transfer Learning for Skin Cancer Segmentation and Classification," 2021 IEEE 21st International Conference on Bioinformatics and Bioengineering (BIBE), Kragujevac, Serbia, 2021, pp. 1-5, doi: 10.1109/BIBE52308.2021.9635175. S. Chen, B. Mulgrew, and P. M.

P. Agrahari, A. Agrawal, and N. Subhashini, “Skin Cancer Detection Using Deep Learning,” Futuristic Communication and Network Technologies, pp. 179–190, Oct. 2021, doi: https://doi.org/10.1007/978-981-16-4625-6_18

M. Upadhyay, J. Rawat, and S. Maji, “Skin cancer image classification using deep neural network models,” in Evolution in Computational Intelligence, Singapore: Springer Nature Singapore, 2022, pp. 451– 460., doi: https://doi.org/10.1007/978-981-16-6616-2_44.

H. C. Reis, V. Turk, K. Khoshelham, and S. Kaya, “InSiNet: a deep convolutional approach to skin cancer detection and segmentation,” Medical & Biological Engineering & Computing, vol. 60, no. 3, pp. 643–662, Jan. 2022, doi: https://doi.org/10.1007/s11517-021-02473-0.

N. Badshah and A. Ahmad, “ResBCU-Net: Deep learning approach for segmentation of skin images,” Biomedical Signal Processing and Control, vol. 71, pp. 103137, Jan. 2022, doi: https://doi.org/10.1016/j.bspc.2021.103137.

ISIC Challenge, ISIC Challenge Databases, 27.10.2023. https://challenge.isic-archive.com/data/

Jaisakthi, S. M., Mirunalini, P. and Aravindan, C., "Automated skin lesion segmentation of dermoscopic images using GrabCut and K-means algorithms," IET Comput. Vis., vol. 12, no. 8, pp. 1088–1095, 2018.

Zhang, Y., & Jiao, R. (2023). Towards Segment Anything Model (SAM) for Medical Image Segmentation: A Survey. http://arxiv.org/abs/2305.03678

A. Kirillov et al., “Segment Anything,” Apr. 2023, Accessed: Oct. 23, 2023. [Online]. Available: https://arxiv.org/abs/2304.02643v1.

W. Ji, J. Li, Q. Bi, T. Liu, W. Li, and L. Cheng, “Segment Anything Is Not Always Perfect: An Investigation of SAM on Different Real-world Applications,” Apr. 2023, Accessed: Oct. 23, 2023. [Online]. Available: https://arxiv.org/abs/2304.05750v3.

C. Mattjie et al., “Zero-shot performance of the Segment Anything Model (SAM) in 2D medical imaging: A comprehensive evaluation and practical guidelines,” Apr. 2023, Accessed: Oct. 23, 2023. [Online]. Available: https://arxiv.org/abs/2305.00109v2.

S. He et al., “Computer-Vision Benchmark Segment-Anything Model (SAM) in Medical Images: Accuracy in 12 Databases,” Apr. 2023, Accessed: Oct. 23, 2023. [Online]. Available: https://arxiv.org/abs/2304.09324v3.

Y. Huang et al., “Segment Anything Model for Medical Images?” Apr. 2023, Accessed: Oct. 25, 2023. [Online]. Available: https://arxiv.org/abs/2304.14660v4.

J. Wu et al., “Medical SAM Adapter: Adapting Segment Anything Model for Medical Image Segmentation,” Apr. 2023, Accessed: Oct. 24, 2023. [Online]. Available: https://arxiv.org/abs/2304.12620v6.

M. Hu, Y. Li, and X. Yang, “SkinSAM: Empowering Skin Cancer Segmentation with Segment Anything Model,” Apr. 2023, Accessed: Oct. 24, 2023. [Online]. Available: https://arxiv.org/abs/2304.13973v1.


Article Metrics

Metrics Loading ...

Metrics powered by PLOS ALM

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Selcuk University Journal of Engineering Sciences (SUJES) ISSN:2757-8828

Abstracting and indexing

Index Copernicus International

scholar_logo_64dp.png

Selcuk university journal of engineering sciences (Online)

ICI World of Journals

ResearchBib