Adversarial training with cycle consistency for unsupervised super-resolution in endomicroscopy
| Autor: | Rav��, Daniele, Szczotka, Agnieszka Barbara, Pereira, Stephen P, Vercauteren, Tom |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2019 |
| Předmět: |
FOS: Computer and information sciences
Technology Colon Computer Vision and Pattern Recognition (cs.CV) ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Computer Science - Computer Vision and Pattern Recognition Video Recording Datasets as Topic Computer Science Artificial Intelligence Article Engineering Esophagus Humans Adversarial training Engineering Biomedical Science & Technology Microscopy Confocal Radiology Nuclear Medicine & Medical Imaging Cycle consistency Deep learning Probe-based confocal laser endomicroscopy Unsupervised super-resolution Computer Science Computer Science Interdisciplinary Applications Life Sciences & Biomedicine Algorithms Unsupervised Machine Learning |
| Zdroj: | Medical Image Analysis Ravì, D, Szczotka, A B, Pereira, S P & Vercauteren, T 2019, ' Adversarial training with cycle consistency for unsupervised super-resolution in endomicroscopy ', Medical Image Analysis, vol. 53, pp. 123-131 . https://doi.org/10.1016/j.media.2019.01.011 |
| ISSN: | 1361-8423 1361-8415 |
| DOI: | 10.1016/j.media.2019.01.011 |
| Popis: | In recent years, endomicroscopy has become increasingly used for diagnostic purposes and interventional guidance. It can provide intraoperative aids for real-time tissue characterization and can help to perform visual investigations aimed for example to discover epithelial cancers. Due to physical constraints on the acquisition process, endomicroscopy images, still today have a low number of informative pixels which hampers their quality. Post-processing techniques, such as Super-Resolution (SR), are a potential solution to increase the quality of these images. SR techniques are often supervised, requiring aligned pairs of low-resolution (LR) and high-resolution (HR) images patches to train a model. However, in our domain, the lack of HR images hinders the collection of such pairs and makes supervised training unsuitable. For this reason, we propose an unsupervised SR framework based on an adversarial deep neural network with a physically-inspired cycle consistency, designed to impose some acquisition properties on the super-resolved images. Our framework can exploit HR images, regardless of the domain where they are coming from, to transfer the quality of the HR images to the initial LR images. This property can be particularly useful in all situations where pairs of LR/HR are not available during the training. Our quantitative analysis, validated using a database of 238 endomicroscopy video sequences from 143 patients, shows the ability of the pipeline to produce convincing super-resolved images. A Mean Opinion Score (MOS) study also confirms this quantitative image quality assessment. Accepted for publication on Medical Image Analysis journal |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect