Mardi 5 avril
13:00 à 14:20
Activité
- Mode rediffusion
- Langue
Relève en intelligence et données - Axe santé et étude du vivant
13h - Présentation d'Anthony Bilodeau (en français)
Augmentation de l'accessibilité des tâches d'imagerie hautement paramétrées par approche de bandits
La microscopie optique par fluorescence de super-résolution est un outil essentiel en biologie pour visualiser les structures sous-cellulaires avec une invasivité minimale. Cependant, les défis techniques associés à la microscopie de super-résolution, à savoir le nombre élevé de paramètres d'imagerie et les objectifs d'imagerie conflictuels, constituent souvent une étape limitante dans son application. Nous abordons ici la tâche d'optimisation multi-objectifs et multi-paramètres par une approche de bandit. Nous évaluons l'approche proposée à l'aide d'un outil de simulation et montrons son applicabilité dans des expériences réelles de microscopie.
13h20 - Présentation de Rupali Bhati
Interpret Your Care : Predicting the Evolution of Symptoms for Cancer Patients
Cancer treatment causes many side-effects like pain, fatigue, etc. Therefore, cancer rehabilitation is becoming increasingly relevant. In this work, we use patient data to predict the evolution of their symptoms such that we can proactively identify adverse effects of cancer to intervene early. We make these predictions using an interpretable machine learning model based on decision trees. We perform experiments on real-world patient data consisting of 20163 patients and provide insights into key features for prediction.
13h40 - Présentation de Rogia Kpanou (en anglais)
On the Robustness of Generalization of Drug-Drug Interaction Models
Drug discovery is a relatively time-consuming and expensive process. Due to drug-drug interactions (DDIs), almost half of the molecules discovered and developed are withdrawn from the market. We present here an in-depth and realistic evaluation of different state-of-the-art algorithms. Our results show that the prediction of DDIs amongst new molecules is extremely challenging. A change in the experiment design is also needed to improve the drug development process.
14h - Présentation de Theresa Wiesner (en anglais)
A Quantitative Analysis of Miniature Synaptic Calcium Transients using Positive Unlabeled Deep Learning
Localized calcium fluctuations within neuronal compartments can be monitored with fluorescence microscopy to improve our understanding of neuronal communication. Current microscopy image analysis methods fail at detecting and segmenting low-intensity miniature synaptic calcium transients (mSCTs). We propose to use a positive/unlabeled deep learning (DL) scheme for quantitative analysis of mSCTs in microscopy time series. The proposed DL framework reduces analysis duration and outperforms the threshold-based analysis baseline for the detection and segmentation of mSCTs.
Anthony Bilodeau | Étudiant au doctorat en biophotonique | Université Laval
Anthony Bilodeau est étudiant au doctorat en biophotonique à l'Université Laval. Il travaille actuellement à son projet projet de doctorat, qui consiste à la construction d'un microscope de super-résolution qui sera en mesure de s'adapter à l'échantillon en utilisant l'apprentissage par renforcement. Pendant sa maîtrise, il a développé des techniques d'apprentissage faiblement supervisées pour faciliter l'analyse d'images de microscopie.
Rupali Bhati | M.Sc. Student | Université Laval
Rupali Bhati is a Master’s in Computer Science student at Université Laval supervised by Audrey Durand. Her research interests include applications of machine learning and reinforcement learning to healthcare and other domains. Prior to her Master’s, she completed her bachelor’s from Delhi Technological University and worked at KPMG.
Rogia Kpanou | On the Robustness of Generalization of Drug-Drug Interaction Models | Université Laval
Rogia kpanou is a doctoral student in computer science. During her master's degree in computer science, she focused on predicting drug-drug interaction side effects using artificial intelligence algorithms. Currently, she is working on how to generate targeted monoclonal antibodies for the treatment of bovine mastitis, again using recent advances in the field of artificial intelligence. In addition to her master's degree in computer science, Rogia kpanou has a bachelor's degree in bioinformatics.
Theresa Wiesner | Stagiaire postdoctorale au Département de psychiatrie et de neurosciences | Université Laval
Theresa Wiesner is a postdoctoral fellow in Biophotonics at Université Laval, where she is working on the development of multimodal super-resolution imaging for live-cell imaging using machine learning. She completed her PhD in Biophotonics under the supervision of Paul De Koninck and Flavie Lavoie-Cardinal in 2021, where she was interested by the quantitative assessment of synaptic plasticity at the molecular scale with multimodal microscopy and computational tools. She completed her master’s degree at Universisté Laval and Université Segalen, in France, and her bachelor degree at the University of Perugia, in Italy.
