University of Twente Student Theses


Segmentation and classification of skin lesions using neural networks

Dokter, J. (2020) Segmentation and classification of skin lesions using neural networks.

[img] PDF
Abstract:When treating diseases involving skin lesions, properly assessing the severity of the disease is crucial in establishing a correct treatment plan. To do this, physicians classify cases based on guidelines. These guidelines often are a trade-off between consistent and quick assessment of the case. This leaves room for improvement on both sides. In order to speed up the assesment process and make it more consistent, the possibili- ties of neural networks are explored. There are two ways in which neural networks are used to analyze images of skin lesion: segmentation and classification. Segmentation is used to detect and to localize the lesion area within the image. It is commonplace in medical image research and has been done on skin lesions before. Classification is used to indicate the severity of several aspects of a lesion or the disease itself. This is much rarer and mostly not in line with existing standards used by physicians, such as the ABCDE score for skin cancer and PASI score for psoriasis. The goal of this research is to explore whether neural networks can be used to classify skin lesions in line with existing medical standards. Segmentation is also used to try and support the classification process. Due to the availability of data, the segmentation will focus on image of lesions regarding skin cancer, while classification focuses on psoriasis lesions and the corresponding PASI score. While these two parts are not directly connected, they explore adjacent applications. In the end, it is examined to see whether these two applications could be used together to improve classification results in the future.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Electrical Engineering MSc (60353)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page