Job Description

In the field of dermatology, the use of deep learning models for skin lesion classification has gained significant traction, offering the potential to improve diagnostic accuracy and efficiency. However, one of the key challenges in developing effective models is the imbalance in available datasets, where certain types of skin lesions are significantly underrepresented. This imbalance can lead to biased models, resulting in lower accuracy, particularly for rare conditions.

Generative Adversarial Networks (GANs) have emerged as a promising solution for generating synthetic data that can help balance these datasets. This project aims to utilize GANs to address the large data imbalance between different diagnoses in our dataset and improve model generalization and accuracy.

Project goals

• Research and evaluate methods for synthetic data generation using GANs
• Develop a proof-of-concept model for synthetic skin lesion data generation
• Evaluate the visual and clinical realism of the generated data
• Analyse and quantify the impact of the synthetic data on model performance, both overall and individual classes.

Qualifications

• Enrolled in a Master's programme in Engineering Physics/Mathematics, Biomedical Engineering, Computer Science, or similar.
• Experience or coursework relating to machine learning and AI projects
• Experience programming in Python and it is a plus if familiar with PyTorch
• Curiosity and interest in the MedTech area.
• Being motivated, creative, focused, and has problem solving skills.

Application and timeframe

Apply with CV, transcript of completed courses, and a brief motivation why you are interested in the project.

The project corresponds to 30hp, starting in January 2025.