Automatic Segmentation of Dermoscopic Images by Iterative Classification
2011

Automatic Segmentation of Dermoscopic Images

Sample size: 122 publication 10 minutes Evidence: moderate

Author Information

Author(s): Zortea Maciel, Skrøvseth Stein Olav, Schopf Thomas R., Kirchesch Herbert M., Godtliebsen Fred

Primary Institution: University of Tromsø

Hypothesis

Can an automatic approach improve the segmentation of skin lesions in dermoscopic images?

Conclusion

The proposed iterative classification segmentation method shows competitive results in accurately segmenting skin lesions compared to existing methods.

Supporting Evidence

  • The proposed method achieved over 91.8% sensitivity in segmenting skin lesions.
  • The algorithm converged in an average of 4.7 iterations.
  • ICS performed similarly well for both benign and malignant lesions.
  • The method was particularly effective for lesions with low contrast.

Takeaway

This study created a computer program that helps doctors find skin cancer by looking at pictures of skin spots. It works by learning from examples and getting better over time.

Methodology

The study used an iterative classification framework that combines initial seed region selection with a hybrid classification strategy to segment skin lesions.

Potential Biases

The accuracy of the segmentation may vary based on the initial assumptions about lesion location and color.

Limitations

The method may not perform well for lesions that cover the entire dermoscopic area or have very low contrast with surrounding skin.

Participant Demographics

The study included 122 dermoscopic images, with 100 benign and 22 malignant lesions.

Digital Object Identifier (DOI)

10.1155/2011/972648

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication