UNCARIA · UNcertainty estimation in CARdiac Image Analysis
„Хоризонт 2020“ — Действия „Мария Склодовска-Кюри“
- Период
- 2021-09-01 → 2024-02-29
- Финансиране от ЕС
- 224 071 €
- Участници
- 2
- Схема
- MSCA-IF
Линиите свързват координатора с партньорите.
Накратко на български
Математически инструменти за изкуствен интелект помагат за по-точно определяне на несигурността при анализ на медицински изображения, например при снимки на сърцето. Това е важно, за да се знае кога моделът е уверен в прогнозата си и кога може да греши.
Кратко обяснение, генерирано от езиков модел по текста на CORDIS. Оригиналът е по-долу.
Резултати накратко
UNCARIA: UNcertainty estimation in CARdiac Image Analysis
- What is the problem being addressed? The overarching goal of the UNCARIA project was to develop mathematical and computational tools and systems for translating medical Machine Learning algorithms from highly accurate, large-scale, pattern recognition systems, into a more clinically usable technology. This is a greatly challenging problem that we have approached by building trustworthiness and transparency on modern DNNs via theoretical frameworks and practical tools for predictive uncertainty analysis. Toward addressing the above research problem, we have developed new techniques to improve model calibration (the ability of a model to be certain when it is correct and uncertain when it tends to be incorrect), and attempt to construct also appropriate measures of success for uncertainty quantification applications. The theoretical approach is complemented by relevant applications: while the initial goal of the project was to focus on cardiac image applications, we have not limited ourselves to this particular area, but have developed methods applicable in most medical image analysis modalities. - Why is it important for society? As Artificial Intelligence adoption grows in different areas of our society, it becomes increasingly important to remain vigilant and aware of its unforeseeable behavior when faced with data far away from its training distribution. In particular uncertainty quantification techniques are useful approaches to measure the true confidence placed by machine learning models on their predictions, which can be critical in medical applications like the ones considered in this project. - What were the overall objectives? The objectives stated next are a slight rectification of the ones initially formulated in the research proposal, as they were been adapted to the continuously changing research landscape of medical machine learning. RO1 – Epistemic Uncertainty in Medical Image Analysis: New tools and methods RO2 – Extension to Quantification of Uncertainty in Medical Image Segmentation RO3– Aleatoric Uncertainty Analysis: Calibration in Medical Image Diagnosis
Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз
Цел на проекта
Cardiovascular diseases account for nearly 45% of all deaths in Europe, with a yearly cost to the EU economy of €210 billions. The emergence of a new generation of deep neural networks (DNNs), powered by higher computing capabilities and the availability of large amounts of data, has enabled unprecedented predictive accuracy, bringing the promise of improving risk assessment and early diagnosis to the field of computational cardiac image understanding. Unfortunately, clinical translation of these tools has not been effectively accomplished yet. A key reason is the black-box nature of these models: through the observation of large-scale annotated data, DNNs can build rich, complex decision boundaries in the image space, but the sequence of mathematical operations leading to such decisions is not readily interpretable by humans.The goal of this project is to open this black-box in a specific direction: building in these models the ability of understanding when they deliver a prediction with a well-founded confidence degree, and when a prediction is reached based only on local statistical regularities of training data and may not be reliable. Current models largely lack this ability, and this undermines their potential for clinical adoption. This project revolves around a fundamental idea: redefining the conventional way of training DNNs so that they can not only produce accurate diagnostic predictions but also model their own errors and have an awareness of them.This proposal involves the transfer of the candidate to a worldwide renowned computer vision group, with a secondment in a top-tier medical research institution, followed by a returning stage in one of the most prestigious biomedical image analysis research groups within Europe. The proposed workplan is designed to train the candidate in both cutting-edge computer vision and clinical knowledge in the outgoing stage, maximizing potential for knowledge transfer to the European host during the incoming phase.
Оригинален текст от CORDIS (на английски).
Участници
Връзки
Данни: CORDIS, © Европейски съюз