Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Extraction and quantification of morphological and spectral features for improving the diagnostic accuracy of skin pathologies

Stasinou Ioanna-Andromachi

Full record


URI: http://purl.tuc.gr/dl/dias/6E8CAD49-0E87-4140-AD07-E898433DB4E3
Year 2022
Type of Item Diploma Work
License
Details
Bibliographic Citation Ioanna-Andromachi Stasinou, "Extraction and quantification of morphological and spectral features for improving the diagnostic accuracy of skin pathologies", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2022 https://doi.org/10.26233/heallink.tuc.94486
Appears in Collections

Summary

Melanoma is one of the most often diagnosed cancers and fatal form of skin cancer. Clinical diagnosis and prognosis of melanoma is a field which remains a challenge for researchers. This study explains in detail the approaches involved for skin cancer diagnostic methods. One of the widely used methods to recognize the skin pathologies and distinguish the melanoma from normal skin is ABCD Rule. Specifically, this study proposes methods based on ABCD Rule, regarding the pre-processing of the input image, skin lesion segmentation and classification of nevus and malignant skin lesions and the extraction and quantification of morphological and spectral features of lesions. To implement this system, a code is performed in MATLAB environment for evaluating the four criteria Asymmetry (A), Border (B), Color (C) and Diameter (D) of a skin lesion. The obtaining features followed by the ABCD Rule are processed by calculating Total Dermoscopic Score (TDS). We deal also with a new type of optical imaging technology, hyperspectral imaging (HSI). As an emerging modality HSI holds great potential for objective cancer assessment and it is a noninvasive high-resolution imaging technique that assists researchers in making more accurate and objective diagnosis of skin cancers. We examine hyperspectral images over the spectral range of 400-880nm and obtain information about melanin content and detect features such as border irregularity, asymmetry and diameter in different wavelengths. In the end of this study, we quote the results which indicate whether our proposed approach is accurate and efficient in the segmentation of the lesion boundary and in the extraction of features of skin lesions and discuss the conclusions extracted for possible future work.

Available Files

Services

Statistics