Detecting failure modes in image reconstructions with interval neural network uncertainty
| Autor: | Jan Macdonald, Wojciech Samek, Maximilian März, Gitta Kutyniok, Luis Oala, Cosmas Heiß |
|---|---|
| Přispěvatelé: | Publica |
| Rok vydání: | 2021 |
| Předmět: |
uncertainty quantification
Computer science 500 Naturwissenschaften und Mathematik::510 Mathematik::510 Mathematik Biomedical Engineering Health Informatics Interval (mathematics) Iterative reconstruction Machine learning computer.software_genre failure modes Image Processing Computer-Assisted Humans Radiology Nuclear Medicine and imaging Uncertainty quantification Artificial neural network business.industry Deep learning Uncertainty deep learning General Medicine Modular design Inverse problem Computer Graphics and Computer-Aided Design Computer Science Applications Image reconstruction Original Article Surgery Neural Networks Computer Computer Vision and Pattern Recognition Artificial intelligence Noise (video) Tomography X-Ray Computed Iímage reconstruction business computer |
| Zdroj: | International Journal of Computer Assisted Radiology and Surgery |
| ISSN: | 1861-6429 1861-6410 |
| Popis: | Purpose The quantitative detection of failure modes is important for making deep neural networks reliable and usable at scale. We consider three examples for common failure modes in image reconstruction and demonstrate the potential of uncertainty quantification as a fine-grained alarm system. Methods We propose a deterministic, modular and lightweight approach called Interval Neural Network (INN) that produces fast and easy to interpret uncertainty scores for deep neural networks. Importantly, INNs can be constructed post hoc for already trained prediction networks. We compare it against state-of-the-art baseline methods (MCDrop, ProbOut). Results We demonstrate on controlled, synthetic inverse problems the capacity of INNs to capture uncertainty due to noise as well as directional error information. On a real-world inverse problem with human CT scans, we can show that INNs produce uncertainty scores which improve the detection of all considered failure modes compared to the baseline methods. Conclusion Interval Neural Networks offer a promising tool to expose weaknesses of deep image reconstruction models and ultimately make them more reliable. The fact that they can be applied post hoc to equip already trained deep neural network models with uncertainty scores makes them particularly interesting for deployment. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink