Keyframe extraction in endoscopic video

Abstract

In medical endoscopy more and more surgeons archive the recorded video streams in a long-term storage. One reason for this development, which is enforced by law in some countries, is to have evidence in case of lawsuits from patients. Another more practical reason is to allow later inspection of previous procedures and also to use parts of such videos for research and for training. However, due to the dramatic amount of video data recorded in a hospital on a daily basis, it is very important to have good preview images for these videos in order to allow for quick filtering of undesired content and for easier browsing through such a video archive. Unfortunately, common shot detection and keyframe extraction methods cannot be used for that video data, because these videos contain unedited and highly similar content, especially in terms of color and texture, and no shot boundaries at all. We propose a new keyframe extraction approach for this special video domain and show that our method is significantly better than a previously proposed approach.

Learning laparoscopic video shot classification for gynecological surgery

“…As a side-effect, the recoded surgery videos benefit the surgeons' work, as they provide a great basis for documentation, training of young surgeons, and medical research. Prior work supporting these aims has been conducted by our research group in the sector of endoscopic video analysis, such as a subjective quality assessment for the impact of compression on the perceived semantic quality [13], instrument classification in laparoscopic videos [17], or extraction and linking of endoscopic key-frames to videos [3,23]. In this work, we restrict ourselves to a very specific field in minimally invasive surgery in the context of gynecology.…”

Automatic workflow for narrow-band laryngeal video stitching

“…The Frame Selection strategy (FS) described in Sec. II-A was compared with standard strategies, as uniform (US) and random (RS) sampling [2]. In these two latter cases, for a fair comparison, the number of extracted frames was kept equal to the number of frames extracted with FS.…”

“…However, the composition of the larynx panorama is a challenging task under different points of view [2]. During the examination, the clinician moves the endoscope inside and outside the larynx, possibly rotating it.…”

Video retrieval in laparoscopic video recordings with dynamic content descriptors

“…Spatio-temporal interest points are often used by numerous motion-based histograms (e.g., Histogram of Oriented Gradients, Histogram of Optical Flow, Motion Boundary Histograms) to represent motion information in a compact way. These methods are hard to use in the medical domain since already the very first step (corner detection) is difficult due to the very special video content, as pointed out by Schoeffmann et al [43] and Primus et al [39]. Therefore, it is unsure how well spatio-temporal keypoints perform for content retrieval in the medical domain (this should be investigated in future work).…”

“…The reasons behind are manifold -among some other motivations the videos are used for: (1) teaching purposes; e.g., to train surgery techniques with inexperienced surgeons, (2) post-operative explanations to the patients, (3) detailed visual information for follow-up surgeries, and (4) evidence in case of lawsuits from patients [34,43]. Another, more recently emerged purpose for usage of the recorded endoscopic video footage is surgical quality assessment (SQA) [10,22].…”