A5077690298- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Explainable Artificial Intelligence (XAI)
- Artificial Intelligence in Healthcare and Education
- Brain Tumor Detection and Classification
- COVID-19 diagnosis using AI
- Medical Image Segmentation Techniques
- Machine Learning and Algorithms
- Cell Image Analysis Techniques
- Machine Learning in Healthcare
- Multiple Sclerosis Research Studies
- Image Processing Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Machine Learning in Materials Science
- Genetics, Bioinformatics, and Biomedical Research
- Adversarial Robustness in Machine Learning
- Viral Infections and Immunology Research
- Machine Learning and Data Classification
- Scientific Computing and Data Management
- Delphi Technique in Research
- Mycobacterium research and diagnosis
- Statistical Methods in Clinical Trials
- Computational Drug Discovery Methods
- Colorectal Cancer Screening and Detection
- Monoclonal and Polyclonal Antibodies Research
Mila - Quebec Artificial Intelligence Institute
2021-2025
McGill University
2021-2025
Intelligent Machines (Sweden)
2021-2022
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...
Discriminative classifiers have become a foundational tool in deep learning for medical imaging, excelling at separable features of complex data distributions. However, these models often need careful design, augmentation, and training techniques to ensure safe reliable deployment. Recently, diffusion synonymous with generative modeling 2D. These showcase robustness across range tasks including natural image classification, where classification is performed by comparing reconstruction errors...
Strong empirical evidence that one machine-learning algorithm A outperforms another B ideally calls for multiple trials optimizing the learning pipeline over sources of variation such as data sampling, augmentation, parameter initialization, and hyperparameters choices. This is prohibitively expensive, corners are cut to reach conclusions. We model whole benchmarking process, revealing variance due initialization hyperparameter choice impact markedly results. analyze predominant comparison...
Abstract Disability progression in multiple sclerosis remains resistant to treatment. The absence of a suitable biomarker allow for phase 2 clinical trials presents high barrier drug development. We propose enable short proof-of-concept by increasing statistical power using deep-learning predictive enrichment strategy. Specifically, multi-headed multilayer perceptron is used estimate the conditional average treatment effect (CATE) baseline and imaging features, patients predicted be most...
While the importance of automatic image analysis is continuously increasing, recent meta-research revealed major flaws with respect to algorithm validation. Performance metrics are particularly key for meaningful, objective, and transparent performance assessment validation used algorithms, but relatively little attention has been given practical pitfalls when using specific a task. These typically related (1) disregard inherent metric properties, such as behaviour in presence class...
Validation metrics are key for the reliable tracking of scientific progress and bridging current chasm between artificial intelligence (AI) research its translation into practice. However, increasing evidence shows that particularly in image analysis, often chosen inadequately relation to underlying problem. This could be attributed a lack accessibility metric-related knowledge: While taking account individual strengths, weaknesses, limitations validation is critical prerequisite making...
The integration of diverse clinical modalities such as medical imaging and the tabular data obtained by patients' Electronic Health Records (EHRs) is a crucial aspect modern healthcare. integrative analysis multiple sources can provide comprehensive understanding patient's condition enhance diagnoses treatment decisions. Deep Neural Networks (DNNs) consistently showcase outstanding performance in wide range multimodal tasks domain. However, complex endeavor effectively merging with clinical,...
Deep learning classifiers are prone to latching onto dominant confounders present in a dataset rather than on the causal markers associated with target class, leading poor generalization and biased predictions. Although explainability via counterfactual image generation has been successful at exposing problem, bias mitigation strategies that permit accurate presence of diverse artifacts remain unsolved. In this work, we propose DeCoDEx framework show how an external, pre-trained binary...
Segmentation of enhancing tumours or lesions from MRI is important for detecting new disease activity in many clinical contexts. However, accurate segmentation requires the inclusion medical images (e.g., T1 post contrast MRI) acquired after injecting patients with a agent Gadolinium), process no longer thought to be safe. Although number modality-agnostic networks have been developed over past few years, they met limited success context pathology segmentation. In this work, we present...
Generalization is an important attribute of machine learning models, particularly for those that are to be deployed in a medical context, where unreliable predictions can have real world consequences. While the failure models generalize across datasets typically attributed mismatch data distributions, performance gaps often consequence biases "ground-truth" label annotations. This context image segmentation pathological structures (e.g. lesions), annotation process much more subjective, and...
There are many clinical contexts which require accurate detection and segmentation of all focal pathologies (e.g. lesions, tumours) in patient images. In cases where there a mix small large standard binary cross entropy loss will result better lesions at the expense missing ones. Adjusting operating point to accurately detect generally leads oversegmentation lesions. this work, we propose novel reweighing strategy eliminate performance gap, increasing pathology while maintaining accuracy. We...
Abstract Progressive forms of multiple sclerosis (MS) remain resistant to treatment. Since there are currently no suitable biomarkers allow for phase 2 trials, pharmaceutical companies must proceed directly financially risky 3 presenting a high barrier drug development. We address this problem through predictive enrichment, which randomizes individuals predicted be most responsive in order increase study’s power. Specifically, deep learning is used estimate conditional average treatment...
Deep learning models can perform well in complex medical imaging classification tasks, even when basing their conclusions on spurious correlations (i.e. confounders), should they be prevalent the training dataset, rather than causal image markers of interest. This would thereby limit ability to generalize across population. Explainability based counterfactual generation used expose confounders but does not provide a strategy mitigate bias. In this work, we introduce first end-to-end...
While deep learning models have achieved remarkable success across a range of medical image analysis tasks, deployment these in real clinical contexts requires that they be robust to variability the acquired images. many methods apply predefined transformations augment training data enhance test-time robustness, may not ensure model's robustness diverse seen patient In this paper, we introduce novel three-stage approach based on transformers coupled with conditional diffusion models, goal...
Precision medicine for chronic diseases such as multiple sclerosis (MS) involves choosing a treatment which best balances efficacy and side effects/preferences individual patients. Making this choice early possible is important, delays in finding an effective therapy can lead to irreversible disability accrual. To end, we present the first deep neural network model individualized decisions from baseline magnetic resonance imaging (MRI) (with clinical information if available) MS Our (a)...
Large, annotated datasets are not widely available in medical image analysis due to the prohibitive time, costs, and challenges associated with labelling large datasets. Unlabelled easier obtain, many contexts, it would be feasible for an expert provide labels a small subset of images. This work presents information-theoretic active learning framework that guides optimal selection images from unlabelled pool labeled based on maximizing expected information gain (EIG) evaluation dataset....
The discovery of patient-specific imaging markers that are predictive future disease outcomes can help us better understand individual-level heterogeneity evolution. In fact, deep learning models provide data-driven personalized much more likely to be adopted in medical practice. this work, we demonstrate biomarker achieved through a counterfactual synthesis process. We show how conditional generative model used perturb local features baseline images pertinent subject-specific evolution and...