Background:

Basal cell carcinoma (BCC) is the most common type of skin cancer. BCCs are classified in three morphological subtypes – superficial, nodular, and infiltrative – which are associated with different outcomes and require different treatment strategies. The morphological classification of BCCs currently relies on pathological analyses. However, this examination involves an invasive tissue sampling procedure, which could be avoided in mild forms of the disease while the more malignant forms require a more extensive resection.

Line-field confocal optical coherence tomography (LC-OCT) is a novel medical imaging technology providing high-resolution 3D acquisitions of the skin layers up to a depth of 400-450 µm. This technology is currently investigated as an alternative for the determination of the BCC morphological subtype at Hôpital Erasme but still needs to be validated to this end. The major limitation of this approach lies in the extraction of relevant criteria from LC-OCT images for classification. To this extent, deep convolutional neural networks – a class of deep learning algorithms – may play a valuable role.

This master thesis project aims at investigating the ability of state-of-the-art deep learning algorithms to predict the morphological subtype in LC-OCT images of BCCs based on pathological supervision.

Contact:

Corentin Martens (corentin.martens@ulb.ac.be)

Gaetan Van Simaeys (gaetan.vansimaeys@ulb.ac.be)