Alice Gosselin

– Image data analysis using mutliplex immunofluorescence data. – Objective: use hierarchical clustering based on embeddings or latent space from autoencoder, in order to help predict drug response, prognosis. – Find patterns/colocalisations of cell types to help understanding cancer progression, response/non-response to a treatment. – Publication review on: – embeddings – deep learning in digital pathology – PhenoGraph – SpatialSort using Bayes Net and Hidden Markov Random Field – Graph Convolutional Neural Network – Self-supervised learning – Training on: image segmentation, image denoising, introduction to Human Pose Estimation, FIJI.

– Oncology Clinical Studies, Biostatistics and Programming Department: clinical trial designs, interactions with clinicians (for designs and health regulatories decisions regarding project development plan), surveillance of competitive environment, Protocols and Statistical Analysis plans – Data analyses (R and Python): – Predictive analyses based on Bayesian methodology – Trends identification for decision making – Dashboards – Understand database structure – Error rate control – Understand bias sources – Population adequacy – Statistical Methodology for data analyses: – Bayesian methods (posterior probabilities, predictive probabilities of success, decision-making tools) – Adaptive Designs for Clinical trials (dose selection, sample-size re-estimation) – Basket Designs – Historical Control Database with innovative proposals such as Synthetic Control Arm framework – Data visualization tool through R shiny application

– Operational activities in a clinical study: – Protocol – Case Report Form – Statistical Analysis Plan – KRM – Data Review – Methodology research on missing data controlled-based imputation: – Joint Modeling Approach: Copy Increment from Reference, Copy Reference, Jump to Reference – Sequential Modeling Approach