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

Search

Browse

My Space

Recognition of nevi on human body in internet images

Syrigos Dimitrios

Full record


URI: http://purl.tuc.gr/dl/dias/1C6ED1F4-7A8D-4C3C-8F0A-5BA2AB8D9BAF
Year 2021
Type of Item Diploma Work
License
Details
Bibliographic Citation Dimitrios Syrigos, "Recognition of nevi on human body in internet images", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.88433
Appears in Collections

Summary

Skin cancer is one of the deadliest forms of cancer. After it metastasizes from its origin into other tissues, the response rate to treatment declines as low as 5%, and its 10-year survival rate is only about 10%. After metastasis, there is no surgical removal option available for treatment.However, an early diagnosis and a surgery removal, significantly increase the probability of survival. Dermoscopy is a noninvasive high-resolution imaging technique that assists physicians in making more accurate diagnoses of skin cancers. Therefore, this thesis proposes highly accurate methods, from three different approaches, regarding the skin lesion segmentation (i.e., isolating the lesion from the rest of the image) and classification of nevus and malignant skin lesions. The main point is to build a system that will be able to identify potentially dangerous cases. We explored through relevant datasets, the effectiveness of both pre-trained and scratch built models with and without segmented images, where the skin lesion area has been isolated as well as with and without cross validation methods. In the end, the results obtained from all these classifiers and approaches are also compared. The study showed that the implementation of Deep learning within the field of cancer diseases can be the most suitable way to classify and recognize skin cancer images, which can be very beneficial in the field of medicine for early diagnosis and improve the accurate diagnosis result. This current work showed an output result of 91% accuracy.

Available Files

Services

Statistics