A5038749016- Medical Image Segmentation Techniques
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Brain Tumor Detection and Classification
- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Artificial Intelligence in Healthcare and Education
- Advanced Neural Network Applications
- Face recognition and analysis
- Medical Imaging Techniques and Applications
- Face and Expression Recognition
- Cell Image Analysis Techniques
- Multiple Sclerosis Research Studies
- Gene expression and cancer classification
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Image Retrieval and Classification Techniques
- Advanced MRI Techniques and Applications
- Explainable Artificial Intelligence (XAI)
- COVID-19 diagnosis using AI
- Domain Adaptation and Few-Shot Learning
- Medical Imaging and Analysis
- Adversarial Robustness in Machine Learning
- Ultrasound Imaging and Elastography
- Machine Learning and Algorithms
McGill University
2016-2025
Mila - Quebec Artificial Intelligence Institute
2021-2025
Centre Universitaire de Mila
2024
Intelligent Machines (Sweden)
2014-2022
Canadian Institute for Advanced Research
2021
Tel Aviv University
2017-2019
Montreal Neurological Institute and Hospital
2001-2016
EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform fair and meaningful comparison registration algorithms which are applied to database intrapatient thoracic CT image pairs. Evaluation nonrigid techniques nontrivial task. This compounded by the fact that researchers typically test only on their own data, varies widely. For this reason, reliable assessment different has been virtually impossible in past. In work we present results launch phase...
International challenges have become the standard for validation of biomedical image analysis methods. Given their scientific impact, it is surprising that a critical common practices related to organization has not yet been performed. In this paper, we present comprehensive conducted up now. We demonstrate importance and show lack quality control consequences. First, reproducibility interpretation results often hampered as only fraction relevant information typically provided. Second, rank...
The Segment Anything Model (SAM) has recently gained popularity in the field of image segmentation due to its impressive capabilities various tasks and prompt-based interface. However, recent studies individual experiments have shown that SAM underperforms medical segmentation, since lack specific knowledge. This raises question how enhance SAM's capability for images. In this paper, instead fine-tuning model, we propose Medical Adapter (Med-SA), which incorporates domain-specific knowledge...
Abstract The Sørensen-Dice index (SDI) is a widely used measure for evaluating medical image segmentation algorithms. It offers standardized of accuracy which has proven useful. However, it diminishing insight when the number objects unknown, such as in white matter lesion multiple sclerosis (MS) patients. We present refinement finer grained parsing SDI results situations where unknown. explore these ideas with two case studies showing what can be learned from our presented studies. Our...
Studies have demonstrated the feasibility of late Gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging for guiding management patients with sequelae to myocardial infarction, such as ventricular tachycardia and heart failure. Clinical implementation these developments necessitates a reproducible reliable segmentation infarcted regions. It is challenging compare new algorithms infarct in left ventricle (LV) existing algorithms. Benchmarking datasets evaluation...
The number of biomedical image analysis challenges organized per year is steadily increasing. These international competitions have the purpose benchmarking algorithms on common data sets, typically to identify best method for a given problem. Recent research, however, revealed that practice related challenge reporting does not allow adequate interpretation and reproducibility results. To address discrepancy between impact quality (control), Biomedical Image Analysis ChallengeS (BIAS)...
This paper presents a novel framework for detecting, localizing, and classifying faces in terms of visual traits, e.g., sex or age, from arbitrary viewpoints the presence occlusion. All three tasks are embedded general viewpoint-invariant model object class appearance derived local scale-invariant features, where features probabilistically quantified their occurrence, appearance, geometry, association with traits interest. An is first learned class, after which Bayesian classifier trained to...
Increasing evidence shows that flaws in machine learning (ML) algorithm validation are an underestimated global problem. Particularly automatic biomedical image analysis, chosen performance metrics often do not reflect the domain interest, thus failing to adequately measure scientific progress and hindering translation of ML techniques into practice. To overcome this, our large international expert consortium created Metrics Reloaded, a comprehensive framework guiding researchers...
Deep learning (DL) models have provided state-of-the-art performance in various medical imaging benchmarking challenges, including the Brain Tumor Segmentation (BraTS) challenges. However, task of focal pathology multi-compartment segmentation (e.g., tumor and lesion sub-regions) is particularly challenging, potential errors hinder translating DL into clinical workflows. Quantifying reliability model predictions form uncertainties could enable review most uncertain regions, thereby building...
In this paper, we propose a new multi-scale technique for multi-modal image registration based on the alignment of selected gradient orientations reduced uncertainty. We show how robustness and accuracy can be improved by restricting evaluation orientation to locations where uncertainty fixed is minimal, which formally demonstrate correspond high magnitude. also embed computationally efficient estimating transformed moving (rather than resampling pixel intensities recomputing gradients)....
Detection of new Multiple Sclerosis (MS) lesions on magnetic resonance imaging (MRI) is important as a marker disease activity and potential surrogate for relapses. We propose an approach where sequential scans are jointly segmented, to provide temporally consistent tissue segmentation while remaining sensitive newly appearing lesions. The method uses two-stage classification process: 1) Bayesian classifier provides probabilistic brain at each voxel reference follow-up scans, 2)...