Video Retrieval System for Meniscal Surgery to Improve Health Care Services

Table 1

Comparison of proposed methodology with existing ones.

Reference papers	Feature set	Feature descriptor	Similarity/cluster/classification measurement	Dataset	Research domain	Video type
Droueche et al. [18]	A motion vector, residual information	Kalman filter, generalized Gaussian distribution	Extension of fast dynamic time warping (EFDTW)	69 video-recorded retinal surgery steps, 1400 video-recorded cataract surgery, and 1707 movie clips with classified human actions	Eye surgery	MPEG-4 AVC/H.264
Ramya et al. [4]	Global features	Entropy, histogram, frequency	Neural network	15478 XML scripts based on key frames	Medical videos	mp4
Schoeffmann et al. [5]	Position, colour, texture	N/A	Earth mover’s distance, signature quadratic form distance, and signature matching distance	25 stored video files per procedure	Laparoscopic surgery	N/A
Beecks et al. [6]	Spatial information (colour, coarseness, contrast value)	Local feature descriptor	-mean cluster	25 segment videos per procedure, resulting in total 1276 short video clips ranging from 60 sec to 250 sec	Endoscopic videos	N/A
Carlos et al. [7]	Local and global features, feature fusion	SIMPLE, CEDD, SURF	-mean	1276	Endoscopy videos	N/A
Charriere et al. [8]	N/A	BoVW	-nearest neighbour (KNN), motion histogram, and hidden Markov model	30	Cataract surgery	DV format
Muhammad et al. [9]	Colour opponent colour space (CoC), curvature map, texture	N/A	N/A	10	Hysteroscopy	N/A
Quellec et al. [10]	Texture, colour, optical flow	N/A	N/A	23 retinal, 100 cataract surgery, 69 hollywood movie	Eye surgery	MPEG 2
Droueche et al. [18]	Motion histogram	Fast dynamic time warping	-mean, Kalman filter, and extended of fast dynamic time warping	23 epiretinal surgery video	Epiretinal membrane surgery	MPEG-4 AVC/H.264
Proposed	The key point, colour	SIFT, HOG, DT, CFD, PCA-SIFT, spatiotemporal feature vector	GMM, Euclidean distance, and Kullback-Leibler divergence	20 meniscus surgery	Meniscus surgery	Flv, Avi, mp4