Taxonomic indexes for differentiating malignancy of lung nodules on CT images

Silva, Giovanni Lucca França da; Carvalho Filho, Antonio Oseas de; Silva, Aristófanes Corrêa; Paiva, Anselmo Cardoso de; Gattass, Marcelo

doi:10.1590/2446-4740.04615

Acessibilidade / Reportar erro

Brasil

Research on Biomedical Engineering

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

Original Articles • Res. Biomed. Eng. 32 (03) • Jul-Sep 2016 • https://doi.org/10.1590/2446-4740.04615 copy

Taxonomic indexes for differentiating malignancy of lung nodules on CT images

Authorship SCIMAGO INSTITUTIONS RANKINGS

Abstract

Introduction

Lung cancer remains the leading cause of cancer mortality worldwide, with one of the lowest survival rates after diagnosis. Therefore, early detection greatly increases the chances of improving patient survival.

Methods

This study proposes a method for diagnosis of lung nodules in benign and malignant tumors based on image processing and pattern recognition techniques. Taxonomic indexes and phylogenetic trees were used as texture descriptors, and a Support Vector Machine was used for classification.

Results

The proposed method shows promising results for accurate diagnosis of benign and malignant lung tumors, achieving an accuracy of 88.44%, sensitivity of 84.22%, specificity of 90.06% and area under the ROC curve of 0.8714.

Conclusion

The results demonstrate the promising performance of texture extraction techniques by means of taxonomic indexes combined with phylogenetic trees. The proposed method achieves results comparable to those previously published.

Keywords:
Medical image; Lung nodule diagnosis; Texture analysis; Taxonomic indexes

Biological term	Adapted term
Community	Region of interest of the CT image
Species	Maximum number of Hounsfield units (HU) in a region
Richness of species: number of species found in a certain region	Richness of species: number of voxels found in a region
Individuals	Number of voxels of a particular species contained in a region
Relative abundance: number of individuals of a certain species existing in a given area	Number of voxels found a the region, which have the same HU value (species)

Experiments	Proportion of the training and test sets	Mean accuracy	Mean sensitivity	Mean specificity	Mean AUC
Tree 1	20 / 80	82.86%	71.82%	87.02%	0.7941
	40 / 60	84.06%	80.84%	85.34%	0.8308
	60 / 40	84.76%	81.62%	85.92%	0.8379
	80 / 20	83.50%	81.10%	84.56%	0.8283
	Standard deviation	0.81	4.69	1.04	-
Tree 2	20 / 80	87.04%	82.24%	88.96%	0.8559
	40 / 60	86.52%	82.88%	88.00%	0.8544
	60 / 40	88.24%	84.30%	89.80%	0.8703
	80 / 20	88.44%	84.22%	90.06%	0.8714
	Standard deviation	0.93	1.02	0.93	-
Tree 3	20 / 80	82.82%	77.08%	85.10%	0.8110
	40 / 60	84.08%	81.08%	85.22%	0.8315
	60 / 40	84.04%	82.28%	84.72%	0.8348
	80 / 20	84.00%	83.40%	84.16%	0.8381
	Standard deviation	0.61	2.75	0.48	-
All trees	20 / 80	86.00%	82.64%	87.34%	0.8500
	40 / 60	87.66%	85.40%	88.50%	0.8694
	60 / 40	87.58%	84.36%	88.76%	0.8658
	80 / 20	88.18%	84.26%	89.88%	0.8708
	Standard deviation	0.94	1.14	1.04	-

#Exp	Tree 1		Tree 2		Tree 3		All Tree
#Exp	C	γ	C	γ	C	γ	C	γ
1	128	0.0078125	0.5	0.125	8	0.125	32	0.0078125
2	32768	0.000488281	32	0.0078125	32	0.125	2	0.03125
3	512	0.0078125	32	0.0078125	2048	0.0078125	8	0.0078125
4	32	0.03125	2	0.5	8	0.125	32	0.0078125
5	126	0.0078125	0.5	0.125	8	0.5	32	0.0078125

Work	Techniques	Database	#Samples	Accuracy	Sensitivity	Specificity
Nascimento et al. (2012)Nascimento LB, Paiva AC, Silva AC. Lung nodules classification in CT images using Shannon and Simpson diversity indices and SVM. In: Perner P. Machine learning and data mining in pattern recognition. Heidelberg: Springer Berlin Heidelberg; 2012. p. 454-66.	Texture features using Diversity Indexes of Shannon and Simpson, Linear Discriminant Analysis and SVM	LIDC	73	92.78%	85.64%	97.89%
Orozco et al. (2013)Orozco MH, Villegas OOV, Dominguez HJO, Sanchez VGC. Lung nodule classification in CT thorax images using support vector machines. In: Proceedings of the 12th Mexican International Conference on Artificial Intelligence (MICAI); 2013; Mexico City, Mexico. USA: IEEE; 2013. p. 277-83.	Texture features, Correlation-based Feature Selection and k-Nearest Neighbor, and SVM	NBIA-ELCAP	113	-	96.15%	52.17%
Krewer et al. (2013)Krewer H, Geiger B, Hall LO, Goldgof DB, Gu Y, Tockman M, Gillies RJ. Effect of texture features in computer aided diagnosis of pulmonary nodules in low-dose computed tomography. In: Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics (SMC); 2013; Manchester, United Kingdom. USA: IEEE; 2013. p. 3887-91.	Texture features, Correlation-based Feature Selection and k-Nearest Neighbor	LIDC-IDRI	33	90.91%	85.71%	94.74%
Dandil et al. (2014)Dandil E, Cakiroglu M, Eksi Z, Ozkan M, Kurt OK, Canan A. Artificial neural network-based classification system for lung nodules on computed tomography scans. In: Proceedings of the 6th International Conference of Soft Computing and Pattern Recognition (SoCPaR); 2014; Tunis, Tunisia. USA: IEEE Transaction on Pattern Analysis and Machine Intelligence; 2014. p. 382-6.	Texture features using Matrix Co-occurrence of Gray Levels, Principal Component Analysis and Artificial Neural Network	Private	128	90.63%	92.30%	89.47%
Parveen and Kavitha (2014)Parveen SS, Kavitha C. Classification of lung cancer nodules using SVM Kernels. International Journal of Computers and Applications. 2014; 95(25):25-8. http://dx.doi.org/10.5120/16751-7013. http://dx.doi.org/10.5120/16751-7013...	Texture features using Matrix Co-occurrence of Gray Levels and SVM	Private	3278	-	91.38%	89.56%
Kuruvilla and Gunavathi (2014)Kuruvilla J, Gunavathi K. Lung cancer classification using neural networks for CT images. Computer Methods and Programs in Biomedicine. 2014; 113(1):202-9. http://dx.doi.org/10.1016/j.cmpb.2013.10.011. PMid:24199657. http://dx.doi.org/10.1016/j.cmpb.2013.10...	Texture features using Statistical Parameters and Artificial Neural Network	LIDC	110	93.30%	91.40%	100%
Gupta and Tiwari (2014)Gupta B, Tiwari S. Lung cancer detection using curvelet transform and neural network. International Journal of Computers and Applications. 2014; 86(1):15-7. http://dx.doi.org/10.5120/14949-3082. http://dx.doi.org/10.5120/14949-3082...	Curvelet Transfom and Artificial Neural Network	Private	120	90.00%	86.66%	93.33%
Hua et al. (2015)Hua K-L, Hsu C-H, Hidayati SC, Cheng WH, Chen YJ. Computer-aided classification of lung nodules on computed tomography images via deep learning technique. OncoTargets and Therapy. 2015; 8:2015-22. http://dx.doi.org/10.2147/OTT.S80733. PMid:26346558. http://dx.doi.org/10.2147/OTT.S80733...	Deep Learning Techniques (Deep Belief Network and Convolutional Neural Network)	LIDC	2545	-	73.30%	78.70%
Kumar et al. (2015)Kumar D, Wong A, Clausi DA. Lung nodule classification using deep features in CT images. In: Proceedings of the 12th Conference on Computer and Robot Vision (CRV); 2015; Halifax, Nova Scotia. USA: IEEE; 2015. p. 133-8.	Deep Learning Technique (Stacked Autoencoder)	LIDC	4323	75.01%	83.35%	-
Tree 1	*Texture features using Taxonomic Indexes (∆ and ∆), Bounding Areas, Phylogenetic Trees and SVM**	LIDC-IDRI	1405	84.76%	81.62%	85.92%
Tree 2		LIDC-IDRI	1405	88.44%	84.22%	90.06%
Tree 3		LIDC-IDRI	1405	84.08%	81.08%	85.22%
All trees together		LIDC-IDRI	1405	88.18%	84.26%	89.88%

Sociedade Brasileira de Engenharia Biomédica Centro de Tecnologia, bloco H, sala 327 - Cidade Universitária, 21941-914 Rio de Janeiro RJ Brasil, Tel./Fax: (55 21)2562-8591 - Rio de Janeiro - RJ - Brazil
E-mail: rbe@rbejournal.org

Acompanhe os números deste periódico no seu leitor de RSS

[1] *e-mail: giovanni-lucca@live.com