A5035530787- Fetal and Pediatric Neurological Disorders
- Domain Adaptation and Few-Shot Learning
- Cancer-related molecular mechanisms research
- Advanced Neural Network Applications
- Adversarial Robustness in Machine Learning
- Stochastic Gradient Optimization Techniques
- Advanced Neuroimaging Techniques and Applications
- Brain Metastases and Treatment
- Radiopharmaceutical Chemistry and Applications
- Machine Learning and Algorithms
- Neonatal and fetal brain pathology
- COVID-19 diagnosis using AI
- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Lung Cancer Diagnosis and Treatment
King's College London
2022-2024
KU Leuven
2015-2024
Technical University of Munich
2022
University Children's Hospital Zurich
2022
University of Zurich
2022
University Hospital of Zurich
2022
Medical University of Vienna
2022
London Women's Clinic
2022
University College London
2022
Tanta University Hospital
2022
Deep learning models for medical image segmentation can fail unexpectedly and spectacularly pathological cases images acquired at different centers than training images, with labeling errors that violate expert knowledge. Such undermine the trustworthiness of deep segmentation. Mechanisms detecting correcting such failures are essential safely translating this technology into clinics likely to be a requirement future regulations on artificial intelligence (AI). In work, we propose...
Abstract Objectives Only few published artificial intelligence (AI) studies for COVID-19 imaging have been externally validated. Assessing the generalizability of developed models is essential, especially when considering clinical implementation. We report development International Consortium Imaging AI (ICOVAI) model and perform independent external validation. Methods The ICOVAI was using multicenter data ( n = 1286 CT scans) to quantify disease extent assess likelihood Reporting Data...
Limiting failures of machine learning systems is paramount importance for safety-critical applications. In order to improve the robustness systems, Distributionally Robust Optimization (DRO) has been proposed as a generalization Empirical Risk Minimization (ERM). However, its use in deep severely restricted due relative inefficiency optimizers available DRO comparison wide-spread variants Stochastic Gradient Descent (SGD) ERM.<br>We propose SGD with hardness weighted sampling,...
Deep learning models for medical image segmentation can fail unexpectedly and spectacularly pathological cases images acquired at different centers than training images, with labeling errors that violate expert knowledge. Such undermine the trustworthiness of deep segmentation. Mechanisms detecting correcting such failures are essential safely translating this technology into clinics likely to be a requirement future regulations on artificial intelligence (AI). In work, we propose...
Limiting failures of machine learning systems is paramount importance for safety-critical applications. In order to improve the robustness systems, Distributionally Robust Optimization (DRO) has been proposed as a generalization Empirical Risk Minimization (ERM). However, its use in deep severely restricted due relative inefficiency optimizers available DRO comparison wide-spread variants Stochastic Gradient Descent (SGD) ERM. We propose SGD with hardness weighted sampling, principled and...