Computer Aided Diagnostic Support System for Skin Cancer: A Review of Techniques and Algorithms

Table 5

Summary of classification performance of some skin cancer detection methods.

Source	year	No. of features selected	Classifier	Total images	Melanoma %	Dysplastic nevi %	Benign	Sens %	Spec %	Accuracy
[249]	1993		CART	353	62		38	94	88
[49]	1994		CART	404	59		41	90	88	80
[121]	1994	22	D.A.	164	11		89	88	89
[255]	1994		ANN	200	40	30	30	95	88
[129]	1994	14	ANN	240; 216	50; 50	16.7	33.3; 50	79.5; 86	86.3; 85.5	82.9; 85.7
[261]	1997		Logistic regression	170	44		56	93	67
[262]	1998	22	Discriminant analysis	917	7		93	93	95
[257]	1998	16	ANN	120	32.5	48.4	27.5	90	74
[263]	1999	26	Discriminant analysis	383	4.7		95.3	100	92
[258]	1999	26	ANN	44	43.2		56.8	97.7	100
[117]	1999	13	Discriminant analysis	147	38.8		61.2	88	81	85
[45]	2000	38	ANN	315	13.3		86.7	92.9	97.8
[159]	2001	21	kNN	5363	1.8	18.8	79.4	87	92
[211]	2001	13	Linear classification	246	25.6	45.1	29.2	100	85
[41]	2002	13	ANN	588	36.9		63.1			94
[246]	2002	10	ANN	147	38.8		61.2	93	92.8
[80]	2003	1	Linear classifier	100	50		50	78	90
[193]	2003	38	LDA + kNN + decision tree	152	27.6		72.4	81	74
[163]	2004	10	Linear classifier; KNN	840	46.5		53.5	95; 98	78; 79
[77]	2004		SVM (third degree polynomial)	977	5.12		94.88	96.4	87.16
[40]	2005	NR	SVM	477	8.8		91.2	84	72
[248]	2005	20	ANN; Discriminant analysis	34	41	59		86; 79	100; 90
[259]	2006	200	SVM	22	45		65			70
[87]	2006	28	Decision tree	224	51.8		48.2	51	97
[237]	2006	3	LR+ multivariate model + ROC	132	17.4		82.6	60.9	95.4	89.4
[118]	2007	18	SVM	564	15.6		84.4	93.3	92.3
[239]	2007	2	Logistic regression (LR)	260	17.7	18.1	64.2	91.3	81–91
[200]	2008	10	Multiple classifiers (SVM, GML, kNN)	358	37.4	32.96	29.6			75.69
[124]	2011	33	kNN	83	55.4		44.6	60.7	80.5	66.7
[126]	2011	6	K-NN; Bayesian; Multilayered perceptron; Combination of three	98	52		48	76.4; 76.4; 78.4; 78.4	70.21; 85.11; 95.74; 97.87	73.47; 80.61; 86.73; 87.76
[135]	2012	12	Multilayer percentron	102	50		50	70.5	87.5	76