Mariana Pinheiro Bento

My name is Mariana Bento, a full-time, tenure-track Assistant Professor in the Department of Biomedical Engineering in the Schulich School of Engineering at the University of Calgary. I have joint appointment in the Department of Electrical and Software Engineering (ESE), where I started my first independent academic position (July 1, 2021), before moving my primary appointment to the newly formed Biomedical Engineering Department (BME) in January 2022. I have masters and doctoral degrees computer engineering at the University of Campinas in Brazil, and postdoctoral fellowship in Radiology and Clinical Neuroscience at UCalgary. My research focus on Artificial Intelligence (AI) for Healthcare applications, considering key aspects like responsibility, trustworthiness, fairness, and bias mitigation. We consider biases related to varying demographics and heterogeneous data (variability related to acquisition protocols). I am the Principal Investigator of the NSERC Discovery Grant Translation of Machine Leaning Methods for Brain Magnetic Resonance Imaging to Heterogeneous Datasets: Bridging the Gap to Generalizable Models (awarded 2021). I currently supervise one high school student, one undergraduate student, four master students, one doctoral student and one postdoctoral fellow.I also developed research considering heterogeneous data, created by combining (or aggregating) datasets from multiple scanners and sites, that I intend to translate to the current proposed project for further analyze stroke. The study and development of AI models with such datasets improve reliability, generalization and, importantly, accelerate their translation to multiple domains. I performed works to standardize heterogeneous datasets, and detect prospective outliers to remove poor-quality samples, evaluating their impact in several ML techniques, leading to a better understanding of data variability. These strategies allow the usage of less complex learning models, which improves the models interpretability. I also developed optimized AI applications across different datasets, evaluating the undesirable variation due to different acquisition parameters, studying the impact of the training protocol when developing AI model using heterogeneous data, and preliminary work on domain adaptation strategies. (Fonte: Currículo Lattes)