Classification of SD-OCT Volumes Using Local Binary Patterns: Experimental Validation for DME Detection

Table 7

Experiment  2—-NN, SVM, RF, and GB classification with BoW for the global and local LBP and local LBP-TOP features with different preprocessing. The optimum number of words was selected based on experiment . The most relevant configurations are indicated with fonts. The configurations whose performance declines with additional preprocessing are indicated with italic font while those with the opposite behavior are indicated with underline font. The highest results which are specified in Table 5 are highlighted in bold.

(a)

Features	Preprocessing	-NN						SVM
		SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%
		Global-LBP	NF	43.7	93.7	43.7	87.5	43.7	62.5	68.7	87.5	62.5	62.5	50.0	56.2
F	43.7		56.2	50.0	75.0	62.5	56.2	56.2	56.2	56.2	75.0	56.2	68.7
FA	56.2		62.5	43.7	81.2	68.7	56.2	56.2	68.7	68.7	68.7	56.2	75.0
Local-LBP	NF	75.0	87.5	50.0	68.7	43.7	43.7	75.0	93.7	50.0	75.0	56.2	56.2
	F	56.2	56.2	50.0	50.0	50.0	43.7	81.2	93.7	68.7	68.7	68.7	75.0
	FA	56.2	43.7	50.0	75.0	50.0	62.5	75.0	93.7	75.0	68.7	68.7	68.7
Local-LBP-TOP	NF	56.2	75.0	56.2	75.0	62.5	56.2	81.2	87.5	75.0	100	56.2	75.0
	F	62.5	43.7	37.5	68.7	43.7	62.5	81.2	81.2	75.0	68.7	81.2	68.7
	F+A	56.2	56.2	68.7	50.0	43.7	62.5	62.5	75.0	68.7	75.0	62.5	81.2

(b)

Features	Preprocessing	RF						GB
		SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%	SE%	SP%
		Global-LBP	NF	68.7	93.7	43.7	62.5	50.0	68.7	56.2	50.0	37.5	31.2	50.0	43.7
F	56.2		50.0	56.2	75.0	50.0	75.0	50.0	56.2	56.2	75.0	43.7	62.5
FA	68.7		50.0	56.2	62.5	62.5	56.2	56.2	50.0	68.7	50.0	43.7	75.0
Local-LBP	NF	81.2	81.2	62.5	56.2	56.2	56.2	75.0	62.5	68.7	87.5	50.0	75.0
	F	56.2	81.2	62.5	68.7	68.7	62.5	68.7	75.0	50.0	75.0	50.0	62.5
	FA	68.7	62.5	62.6	68.7	43.7	43.7	56.2	50.0	68.7	56.2	50.0	50.0
Local-LBP-TOP	NF	68.7	62.5	68.7	81.2	68.7	68.7	37.5	68.7	62.5	81.2	62.5	50.0
	F	50.0	62.5	62.5	62.5	43.7	75.0	50.0	56.2	43.7	62.5	50.0	62.5
	F+A	50.0	62.5	81.2	87.5	50.0	68.7	56.2	62.5	81.2	68.7	75.0	68.7