A5008902493- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Medical Image Segmentation Techniques
- Neural Networks and Applications
- Domain Adaptation and Few-Shot Learning
- Hepatocellular Carcinoma Treatment and Prognosis
- Multimodal Machine Learning Applications
- Brain Tumor Detection and Classification
- Medical Imaging and Analysis
- COVID-19 diagnosis using AI
- Spectroscopy Techniques in Biomedical and Chemical Research
- Generative Adversarial Networks and Image Synthesis
- Parallel Computing and Optimization Techniques
- Sparse and Compressive Sensing Techniques
- Machine Learning and Data Classification
- Blind Source Separation Techniques
- Neural dynamics and brain function
- Medical Imaging Techniques and Applications
- Stochastic Gradient Optimization Techniques
- Advanced Scientific Techniques and Applications
- Explainable Artificial Intelligence (XAI)
- Neural Networks and Reservoir Computing
- Machine Learning and Algorithms
- Biomedical and Engineering Education
Polytechnique Montréal
2014-2022
Mila - Quebec Artificial Intelligence Institute
2022
Université de Montréal
2019
Hôpital Saint-Luc
2017
Centre Hospitalier de l’Université de Montréal
2017
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
International challenges have become the de facto standard for comparative assessment of image analysis algorithms given a specific task. Segmentation is so far most widely investigated medical processing task, but various segmentation typically been organized in isolation, such that algorithm development was driven by need to tackle single clinical problem. We hypothesized method capable performing well on multiple tasks will generalize previously unseen task and potentially outperform...
Semantic segmentation of medical images aims to associate a pixel with label in image without human initialization. The success semantic algorithms is contingent on the availability high-quality imaging data corresponding labels provided by experts. We sought create large collection annotated datasets various clinically relevant anatomies available under open source license facilitate development algorithms. Such resource would allow: 1) objective assessment general-purpose methods through...
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
Abstract The analysis of histopathology images with artificial intelligence aims to enable clinical decision support systems and precision medicine. success such applications depends on the ability model diverse patterns observed in pathology images. To this end, we present Virchow, largest foundation for computational date. In addition evaluation biomarker prediction cell identification, demonstrate that a large enables pan-cancer detection, achieving 0.95 specimen-level area under...
We propose a model for the joint segmentation of liver and lesions in computed tomography (CT) volumes. build from two fully convolutional networks, connected tandem trained together end-to-end. evaluate our approach on 2017 MICCAI Liver Tumour Segmentation Challenge, attaining competitive lesion detection scores across wide range metrics. Unlike other top performing methods, output post-processing is trivial, we do not use data external to challenge, simple single-stage that However, method...
It is well known that it challenging to train deep neural networks and recurrent for tasks exhibit long term dependencies. The vanishing or exploding gradient problem a issue associated with these challenges. One approach addressing gradients use either soft hard constraints on weight matrices so as encourage enforce orthogonality. Orthogonal preserve norm during backpropagation may therefore be desirable property. This paper explores issues optimization convergence, speed stability when...
To evaluate the performance, agreement, and efficiency of a fully convolutional network (FCN) for liver lesion detection segmentation at CT examinations in patients with colorectal metastases (CLMs).This retrospective study evaluated an automated method using FCN that was trained, validated, tested 115, 15, 26 contrast material-enhanced containing 261, 22, 105 lesions, respectively. Manual by radiologist reference standard. Performance user-corrected segmentations compared manual...
The use of artificial intelligence to enable precision medicine and decision support systems through the analysis pathology images has potential revolutionize diagnosis treatment cancer. Such applications will depend on models' abilities capture diverse patterns observed in images. To address this challenge, we present Virchow, a foundation model for computational pathology. Using self-supervised learning empowered by DINOv2 algorithm, Virchow is vision transformer with 632 million...
Modelling long-term dependencies is a challenge for recurrent neural networks. This primarily due to the fact that gradients vanish during training, as sequence length increases. Gradients can be attenuated by transition operators and are or dropped activation functions. Canonical architectures like LSTM alleviate this issue skipping information through memory mechanism. We propose new architecture (Non-saturating Recurrent Unit; NRU) relies on mechanism but forgoes both saturating functions...
Abstract In radiation oncology, predicting patient risk stratification allows specialization of therapy intensification as well selecting between systemic and regional treatments, all which helps to improve outcome quality life. Deep learning offers an advantage over traditional radiomics for medical image processing by salient features from training data originating multiple datasets. However, while their large capacity combine high-level imaging prediction, they lack generalization be used...
Foundation models in computational pathology promise to unlock the development of new clinical decision support systems and for precision medicine. However, there is a mismatch between most analysis, which defined at level one or more whole slide images, foundation date, process thousands image tiles contained separately. The requirement train network aggregate information across large number multiple images limits these models' impact. In this work, we present slide-level model H&E-stained...
Foundation models are rapidly being developed for computational pathology applications. However, it remains an open question which factors most important downstream performance with data scale and diversity, model size, training algorithm all playing a role. In this work, we present the result of scaling both surpassing previous studies in dimensions, introduce two new models: Virchow 2, 632M parameter vision transformer, 2G, 1.85B each trained 3.1M histopathology whole slide images. To...
The early detection, diagnosis and monitoring of liver cancer progression can be achieved with the precise delineation metastatic tumours. However, accurate automated segmentation remains challenging due to presence noise, inhomogeneity high appearance variability malignant tissue. In this paper, we propose an unsupervised tumour framework using a machine learning approach based on discriminant Grassmannian manifolds which learns tumours respect normal First, within-class between-class...
External radiation therapy planning is a highly complex and tedious process as it involves treating large target volumes, prescribing several levels of doses, well avoiding irradiating critical structures such organs at risk close to the tumor target. This requires trained dosimetrists physicists generate personalized plan adapt treatment evolves, thus affecting overall control patient outcomes. Our aim achieve accurate dose predictions for head neck (H&N) cancer patients on challenging...