iVENDIS stands for Intelligent in-Vivo Endoscopic Diagnosis and Intervention Support Systems. Within this project, we aim to provide computer vision tools that can aid endoscopists in two main clinical tasks: guidance to lesions for the case of bronchoscopy image analysis and a prediction of lesion histology for the case of colonoscopy videos.
For the case of colonoscopy videos, the following tools have been developed within the scope of the project:
-Image acquisition module: we have developed a software that can extract high quality frames during endoscopy intervention, offering higher quality than the snapshots that can be gathered using available video processors.
-Image preprocessing module: automatic detection and correction of specular highlights. You can find more information about this in the following publication:
Sánchez, F. J., Bernal, J., Sánchez-Montes, C., de Miguel, C. R., & Fernández-Esparrach, G. (2017). Bright spot regions segmentation and classification for specular highlights detection in colonoscopy videos. Machine Vision and Applications, 28(8), 917-936.
-Real time spatio-temporal stable polyp detection (in collaboration with ENSEA): novel polyp detection method which is specially focused on offering real-time performance and which has been developed with its potential integration into embedded systems for an actual clinical use. You can find more information about this method here:
Angermann, Q., Bernal, J., Sánchez-Montes, C., Hammami, M., Fernández-Esparrach, G., Dray, X., … & Histace, A. (2017). Towards Real-Time Polyp Detection in Colonoscopy Videos: Adapting Still Frame-Based Methodologies for Video Sequences Analysis. In Computer Assisted and Robotic Endoscopy and Clinical Image-Based Procedures (pp. 29-41). Springer, Cham.
– Semantic segmentation of colonoscopy images: we have studied the use of trending techniques such as convolutional neural networks to obtain a first complete segmentation of the different elements that appear in the colonoscopy scene (polyps, specular highlights, luminal region). You can find more information in the following reference:
Vázquez, D., Bernal, J., Sánchez, F. J., Fernández-Esparrach, G., López, A. M., Romero, A., … & Courville, A. (2017). A benchmark for endoluminal scene segmentation of colonoscopy images. Journal of healthcare engineering, 2017.
– in-vivo Histology Prediction: we have developed a first algorithm to accurately predict lesion histology. The final method has just been accepted for publication at Endoscopy journal but you can find a preliminary version of the method here:
Sánchez, F. J., Bernal, J., Sánchez-Montes, C., de Miguel, C. R., & Fernández-Esparrach, G. (2017). Bright spot regions segmentation and classification for specular highlights detection in colonoscopy videos. Machine Vision and Applications, 28(8), 917-936.
Apart from the different methodologies, we have also developed supporting tools for method validation. One of them is GTCreator, a fully flexible annotation tool that has been designed to ease annotation tasks. You can find more information about the tool in the following reference:
Bernal, J., Histace, A., Masana, M., Angermann, Q., Sánchez-Montes, C., de Miguel, C. R., … & Fernández-Esparrach, G. (2018). GTCreator: a flexible annotation tool for image-based datasets. International journal of computer assisted radiology and surgery, 1-11.
Please take a look at the corresponding methodology pages to get more details on each of the tools.