Marlies Wakkee

Erasmus MC Cancer Institute

In short, Marlies Wakkee's research entails the following: 

Finding the needle in the haystack: Using artificial intelligence to identify histogenomic biomarkers for metastatic cutaneous squamous cell carcinoma

Patients with cutaneous squamous cell carcinoma with an aggressive clinical course are now poorly identified because they are rare (only 2% of patients develop metastasis). Therefore, these patients now don’t receive optimal follow-up or miss potentially life-saving adjuvant treatments. It is currently impossible to select patients for clinical trials to investigate these treatments because the best staging system has a positive predictive value of 13%. Therefore, it is essential to find the "needle in the haystack" to better treat this patient group in the future.

To do this, we will look at the excised material from patients with cutaneous squamous cell carcinoma who we already know have developed metastasis and as a control group we will use patients who we know have not developed metastasis. These groups are matched on pre-existing histopathological features so that only new biomarkers are found using artificial intelligence (AI). We will look at the excised material in two ways: 1) using pathologist's annotations (segmentation of tissue structures with a convolutional neural network) 2) hypothesis-free approach (using streaming stochastic gradient descent for whole slide images, state-of-the-art classification architectures, layer-wise relevance propagation combined with concept-whitening - for interpretation of biomarkers). Furthermore, we also have RNA sequencing data and we will see if this in combination with the biomarkers found through the above strategies leads to even better identification of the high-risk squamous cell carcinoma patient.

This study is a collaboration between researchers at Erasmus MC (Marlies Wakkee, Antien Mooyaart, Harmen van de Werken and Loes Hollestein) and Radboud UMC (Avital Amir and Geert Litjens).