Melanoma Detection and Classi fication of Birthmarks Using Neural Networks and Genetic Programing |
Henrik Mygind
|
| Abstract | This is the master project of Henrik Mygind developed in collaboration with Rigshospitalet, Pallas Informatik and the DTU Informatics at The Technical University of Denmark. In this thesis a prototype software is developed which is able to classify digital color images of birthmarks determining if a birthmark is melanoma or harmless. A feature extraction algorithm is developed in Matlab and two classifi cation algorithms are developed in C#. One classification algorithm is build on genetic programing while the other use neural networks trained by backpropagation and a genetic algorithm.
The software is tested using a dataset received from Rigshospitalet containg 164 images of melanoma and 172 images of harmless birthmarks. The tests have shown that the feature extraction is able to fi nd the birthmarks in 64% of the images while the classi fication algorithms works as intended when tested on a general dataset. Using the feature extraction developed for this project and the best classi fication algorithm, a classi fication rate of 78% is received. The classifi cation rate is found when the false positive rate and the false negative rate are equal. When using images on which the feature extraction works perfectly a slightly higher classifi cation rate of 82% is reached.
Even at this early stage in the development the software looks promising and when a better feature extraction algorithm is developed the software can be a huge help in the early detection of melanoma around the world. |
| Type | Master's thesis [Academic thesis] |
| Year | 2011 |
| Publisher | Technical University of Denmark, DTU Informatics, E-mail: reception@imm.dtu.dk |
| Address | Asmussens Alle, Building 305, DK-2800 Kgs. Lyngby, Denmark |
| Series | IMM-M.Sc.-2011-53 |
| Note | Supervised by Associate Professor Carsten Witt, cawi@imm.dtu.dk, DTU Informatics |
| Publication link | http://www.imm.dtu.dk/documents/ftp/ep2011/ep11_53-net.pdf |
| BibTeX data | [bibtex] |
| IMM Group(s) | Computer Science & Engineering |