Object classification using Radar: Use case Melanoma detection

 Thesistype: Bachelor-/Master-/Student Thesis
Contact: Daniel Onwuchekwa - H-E 010, +49271 740 2897
Description: Skin cancer is among the most common types of cancer. It is caused by excessively exposing the skin to ultraviolet radiation usually emitted from the sun. The accuracy of skin cancer diagnosis by dermatology still lies between 75-84%. The accuracy of diagnosis with an unaided eye is much lower at 60%. Although dermoscopy improves the accuracy of melanoma diagnosis in comparison with the inspection by the unaided eye, this accuracy is mainly dependent on the experience of the examiner. Dermoscopy is limited in the diagnosis of very early melanomas, and mainly featureless melanomas. The limitations are because the approach depends on the appearance of classic dermoscopic features. The exploitation of artificial intelligence to assist doctors in non-invasive diagnosis poses great potentials. Based on the premise that the dielectric properties of the healthy skin layer are different from the Melanoma, the mmWave frequency of the electromagnetic spectrum would be emitted from a transmitter towards the skin. The reflective scatter signals of the mmWave upon penetrating the skin would be captured by a receiving antenna. The data received by the antenna is processed and fused with corresponding data from a dermoscopic image in one experiment, and with a regular camera image in a second experiment. The aim is to aid the quick diagnosis of Melanoma.

Task Objectives
• You are required to investigate the existing state of Art for the detection on Melanoma • You are required to investigate existing state or Art Radar sensors • You are required to carry out an experiment to collate data of the skin property using a radar sensor. • Train a deep neural network model for the classification of Melanoma. • Exploit techniques to combine existing detection techniques with the machine learning approach.

Desired Skills


Language
English

Duration of the Thesis
6 Months