WCE polyp detection with triplet based embeddings
Wireless capsule endoscopy is a medical procedure used to visualize the entire gastrointestinal tractand to diagnose intestinal conditions, such as polyps or bleeding. Current analyses are performedby manually inspecting nearly each one of the frames of the video, a tedious and error-prone task.Auto...
| Autores: | , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2020 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:2445/172674 |
| Acceso en línea: | https://hdl.handle.net/2445/172674 |
| Access Level: | acceso abierto |
| Palabra clave: | Aprenentatge automàtic Càpsula endoscòpica Diagnòstic per la imatge Xarxes neuronals convolucionals Pòlips (Patologia) Machine learning Capsule endoscopy Diagnostic imaging Convolutional neural networks Polyps (Pathology) |
| Sumario: | Wireless capsule endoscopy is a medical procedure used to visualize the entire gastrointestinal tractand to diagnose intestinal conditions, such as polyps or bleeding. Current analyses are performedby manually inspecting nearly each one of the frames of the video, a tedious and error-prone task.Automatic image analysis methods can be used to reduce the time needed for physicians to evaluate acapsule endoscopy video. However these methods are still in a research phase.In this paper we focus on computer-aided polyp detection in capsule endoscopy images. This is achallenging problem because of the diversity of polyp appearance, the imbalanced dataset structureand the scarcity of data. We have developed a new polyp computer-aided decision system thatcombines a deep convolutional neural network and metric learning. The key point of the method isthe use of the Triplet Loss function with the aim of improving feature extraction from the imageswhen having small dataset. The Triplet Loss function allows to train robust detectors by forcingimages from the same category to be represented by similar embedding vectors while ensuring thatimages from different categories are represented by dissimilar vectors. Empirical results show ameaningful increase of AUC values compared to state-of-the-art methods.A good performance is not the only requirement when considering the adoption of this technologyto clinical practice. Trust and explainability of decisions are as important as performance. Withthis purpose, we also provide a method to generate visual explanations of the outcome of our polypdetector. These explanations can be used to build a physician's trust in the system and also to conveyinformation about the inner working of the method to the designer for debugging purposes. |
|---|