An Adaptation of Fourier Descriptors’ Spectral Energy for the Classification of Skin Cancer Melanoma

---

Riva C., Turan M., Sezgin M.

TRAITEMENT DU SIGNAL, vol.43, no.1, pp.1-11, 2026 (SCI-Expanded)

• Publication Type: Article / Article
• Volume: 43 Issue: 1
• Publication Date: 2026
• Doi Number: 10.18280/ts.430111
• Journal Name: TRAITEMENT DU SIGNAL
• Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Compendex, zbMATH
• Page Numbers: pp.1-11
• İstanbul Ticaret University Affiliated: Yes

Abstract

Skin cancer has various forms and has widely spread throughout the world, with an

increasing case rate lately. Diagnosis through visual inspection or more enhanced images

can be misleading due to the similarity of mixed cases. Due to the great challenge of this

problem, researchers have attempted to extract different features of lesions and feed them

into machine learning models or directly apply images to deep learning models in order to

increase accuracy. Although they were partially successful, the problem has still not been

fully resolved due to the lack of accuracy in complicated lesions. We propose a novel feature

called slope via energy spectral analysis of Fourier shape descriptors, supported additionally

with lesion features such as average color and a few shape properties. This feature set was

applied to a small-size ANN model for binary classification. The public complicated dataset,

International Skin Imaging Collaboration (ISIC), has been used to train and test the model.

The accuracy was promising, balanced within the classes, and as high as 94%.