Machine Learning Techniques for Quantification of Knee Segmentation from MRI

Table 2

Semiautomated segmentation approaches.

Study	MR sample	Number of subjects	Methods used	Efficiency measures
[63]	PDW SPAIR	12 subjects	Random forest	Femur DSC 94.9%; Tibia DSC 92.5%
[27]	T1-weighted axial	103 subjects	Support vector machine	Accuracy 72%
[58]	CCBR	159 subjects	PLDS method	Dice volume 0.82
[64]	3D SGPR	4 subjects	Region growing	Error −6.53%
[65]	T1-weighted image	5 subjects	Active shape model	Mean error −0.57
[66]	Flash 3D	15 subjects	B-spline with manual adjustment	Interobserver 3.3 to 13.6
[45]	3DMR	20 subjects	Active contour	Thickness 0.996 and 0.998 for femur and tibia
[4]	SPGR	7 subjects	Watershed transformation	DSC 89.5%; Sensitivity 90%; Specificity 99.9%
[23]	3D DESS	320 slices	Graph-cut algorithm	DSC 94.3%
[67]	Flash GRE	50 subjects	K-means with manual adjustment	DSC 0.77 and 0.80; Sensitivity 83.1 and 85.3; Specificity 99.9 and 099.9 for femur and tibia bones, respectively
[68]	DESS	17 subjects	Support vector machine	DSC patella 0.82, tibia 0.83, and femur 0.86
[23]	3T MR images	10 subjects	Graph-cut method	DSC 0.943
[13]	3D DESS	12 subjects	Active contour model	Root mean square 0.8% to 1.3%
[69]	3D DESS	10 subjects	Mesh morphing approach	Mean S.D of femur 0.87, tibia 0.40 and, patella 0.53
[70]	T1-weighted	15 subjects	Watershed method	Cartilage volume tibia 3.3 mm