An Efficient Skin Cancer Diagnostic System Using Bendlet Transform and Support Vector Machine

S, POOVIZHI; TR, GANESH BABU

doi:10.1590/0001-3765202020190554

Acessibilidade / Reportar erro

Brasil

Anais da Academia Brasileira de Ciências

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

BIOMEDICAL SCIENCES • An. Acad. Bras. Ciênc. 92 (01) • 2020 • https://doi.org/10.1590/0001-3765202020190554 copy

An Efficient Skin Cancer Diagnostic System Using Bendlet Transform and Support Vector Machine

Authorship SCIMAGO INSTITUTIONS RANKINGS

Abstract

Skin is the outermost and largest organ of the human body that protects us from the external agents. Among the various types of diseases affecting the skin, melanoma (skin cancer) is the most dangerous and deadliest disease. Though it is one of the dangerous forms of cancer, it has a high survival rate if and only if it is diagnosed at the earliest. In this study, skin cancer classification (SCC) system is developed using dermoscopic images. It is considered as a classification problem with the help of Bendlet Transform (BT) as features and Support Vector Machine (SVM) as a classifier. First, the unwanted information’s such as hair and noises are removed using median filtering approach. Then, directional representation based feature extraction system that precisely classifies curvature, location and orientation is employed. Finally, two SVM classifiers are designed for the classification. The performance of the SCC system based on Bendlet is superior to other image representation systems such as Wavelets, Curvelets, Contourlets and Shearlets.

Key words
Bendlet Transform; Medical image classification; Multi-resolution analysis; Skin cancer; Support vector machine

Performance measure	Description	Formula
Sensitivity (S_n)	It is the ratio between the number of T_rP_o decisions by the classifier to the actual P_o cases tested.	$S_{n} = \frac{T_{r} P_{o}}{T_{r} P_{o} + F_{a} N_{e}}$
Specificity (S_p)	It is the ratio between the number of T_rN_e decisions by the classifier to the actual N_e cases tested.	$S_{p} = \frac{T_{r} N_{e}}{T_{r} N_{e} + F_{a} P_{o}}$
Accuracy (A_c)	It is defined as the number of correctly classified skin cancer images to the total number of images tested.	$A_{c} = \frac{T_{r} P_{o} + T_{r} N_{e}}{T_{r} P_{o} + F_{a} N_{e} + T_{r} N_{e} + F_{a} P_{o}}$
T_rP_o→True Positive, F_aN_e→False Negative, T_rN_e→True Negative and F_aP_o→False Positive

Level of Decomposition	Number of directions	Performance Measures
Level of Decomposition	Number of directions	T_rP_o	F_aN_e	T_rN_e	F_aP_o	S_n (%)	S_p (%)	A_c (%)
1	2	90	30	65	15	75.00	81.25	77.50
	4	94	26	67	13	78.33	83.75	80.50
	8	99	21	68	12	82.50	85.00	83.50
	16	94	26	68	12	78.33	85.00	81.00
	32	86	34	67	13	71.67	83.75	76.50
2	2	97	23	70	10	80.83	87.50	83.50
	4	102	18	72	8	85.00	90.00	87.00
	8	108	12	75	5	90.00	93.75	91.50
	16	100	20	71	9	83.33	88.75	85.50
	32	93	27	71	9	77.50	88.75	82.00
3	2	104	16	77	3	86.67	96.25	90.50
	4	110	10	79	1	91.67	98.75	94.50
	8	117	3	80	0	97.50	100.00	98.50
	16	109	11	77	3	90.83	96.25	93.00
	32	102	18	75	5	85.00	93.75	88.50
4	2	97	23	79	1	80.83	98.75	88.00
	4	108	12	79	1	90.00	98.75	93.50
	8	110	10	79	1	91.67	98.75	94.50
	16	104	16	79	1	86.67	98.75	91.50
	32	96	24	78	2	80.00	97.50	87.00

Level of Decomposition	Number of directions	Performance Measures
Level of Decomposition	Number of directions	T_rP_o	F_aN_e	T_rN_e	F_aP_o	S_n (%)	S_p (%)	A_c (%)
1	2	34	6	67	13	85.00	83.75	84.17
	4	35	5	69	11	87.50	86.25	86.67
	8	36	4	74	6	90.00	92.50	91.67
	16	34	6	72	8	85.00	90.00	88.33
	32	34	6	63	17	85.00	78.75	80.83
2	2	36	4	70	10	90.00	87.50	88.33
	4	37	3	74	6	92.50	92.50	92.50
	8	39	1	77	3	97.50	96.25	96.67
	16	36	4	74	6	90.00	92.50	91.67
	32	33	7	72	8	82.50	90.00	87.50
3	2	38	2	71	9	95.00	88.75	90.83
	4	38	2	77	3	95.00	96.25	95.83
	8	40	0	80	0	100.00	100.00	100.00
	16	37	3	75	5	92.50	93.75	93.33
	32	36	4	72	8	90.00	90.00	90.00
4	2	33	7	74	6	82.50	92.50	89.17
	4	37	3	76	4	92.50	95.00	94.17
	8	39	1	78	2	97.50	97.50	97.50
	16	36	4	75	5	90.00	93.75	92.50
	32	32	8	74	6	80.00	92.50	88.33

Author	#Images used	Accuracy (%)	Sensitivity (%)	Specificity (%)
Ozkan & Koklu (2017)	200	92.50	90.86	96.11
Zaqout (2016)	200	90	85	92.22
Sonia (2016)	120	96.7	97.5	96.3
Nasir et al. (2018)	150	97.50	97.70	96.70
Proposed system	200	98.75	99.25	100

Academia Brasileira de Ciências Rua Anfilófio de Carvalho, 29, 3º andar, 20030-060 Rio de Janeiro RJ Brasil, Tel: +55 21 3907-8100 - Rio de Janeiro - RJ - Brazil
E-mail: aabc@abc.org.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] Correspondence to: Poovizhi S E-mail: poovizhiece22@gmail.com