Fully automatic segmentation of glottis and vocal folds in endoscopic laryngeal high-speed videos using a deep Convolutional LSTM Network - dataset & code

This repository is associated with the manuscript "Fully automatic segmentation of glottis and vocal folds in endoscopic laryngeal high-speed videos using a deep Convolutional LSTM Network", where we used a deep Convolutional Neural Network (CNN) approach for the first time to fully automatically segment not only the time-varying glottal area but also the vocal fold tissue directly from laryngeal HS video. The approach was developed and intensely evaluated on a dataset comprising 130 HS-sequences (13,000 HS video frames in total) obtained from healthy as well as pathologic subjects.

Here, we provide the used dataset, the code along with the best performing Neural Network, and scripts to evaluate the segmentation performance.

This work was supported by the German Research Foundation (DFG), LO-1413/2-2. Computational resources were provided by the High Performance Compute Cluster 'Elwetritsch' at the University of Kaiserslautern, which is part of the 'Alliance of High Performance Computing Rheinland-Pfalz' (AHRP). We kindly acknowledge the support.