Evolutionary Algorithm-Based Classifier Parameter Tuning for Automatic Ovarian Cancer Tissue Characterization and Classification

Abstract

Purpose: Ovarian cancer is one of the most common gynecological cancers in women. It is difficult to accurately and objectively diagnose benign and malignant ovarian tumors using ultrasound and other tests. Hence, there is an imperative need to develop a computer-aided diagnostic (CAD) system for ovarian tumor classification in order to reduce patient anxiety and the cost of unnecessary biopsies. In this paper, we present an automatic CAD system for the detection of benign and malignant ovarian tumors using advanced image processing and data mining techniques.

Materials and Methods: In the proposed system, Hu’s invariant moments, Gabor transform parameters and entropies are first extracted from the acquired ultrasound images. Significant features are then used to train a probabilistic neural network (PNN) classifier for classifying the images into benign and malignant categories. The model parameter (σ) for which the PNN classifier performs the best is identified using a genetic algorithm (GA).

Results: The proposed system was validated using 1300 benign images and 1300 malignant images, obtained from 10 patients with a benign disease and 10 with a malignant disease. We used 23 statistically significant (p < 0.0001) features. By evaluating the classifier using a ten-fold cross-validation technique, we were able to achieve an average classification accuracy of 99.8 %, sensitivity of 99.2 % and specificity of 99.6 % with a σ of 0.264.

Conclusion: The proposed system is automated and hence is more objective, can be easily deployed in any computer, is fast and accurate and can act as an adjunct tool in helping physicians make a confident call about the nature of the ovarian tumor under evaluation.

Zusammenfassung

Ziel: Eierstockkrebs ist eines der häufigsten gynäkologischen Karzinome bei Frauen. Es ist schwierig, durch Sonografie und andere Teste gutartige und bösartige Tumore der Eierstöcke exakt und objektiv zu diagnostizieren. Deshalb ist es zwingend erforderlich, ein computergestütztes diagnostisches (CAD) System für die Klassifikation von Eierstockkarzinomen zu entwickeln, um die Ängste der Patientinnen zu verringern und die von unnötigen Biopsien verursachten Kosten einzusparen. In dieser Publikation stellen wir ein automatisches CAD-System für den Nachweis gutartiger und maligner Eierstockkarzinome vor. Das eine erweiterte Bildverarbeitung und Techniken zur Auswertung größerer Datenmengen (Datamining) verwendet.

Material und Methoden: Im vorgestellten System werden zuerst die Hu‘schen invarianten Momente, die Gabortransformations-Parameter und Entropien aus den erhaltenen Ultraschallbildern extrahiert. Signifikante Merkmale werden dann verwendet, um einen neuralen Wahrscheinlichkeitsnetzwerks(PNN)-Klassifikator zu trainieren, der die Darstellungen in benigne und maligne Kategorien einteilt. Der Modellparameter (σ), für den der PNN-Klassifikator die besten Ergebnisse erzielt, wird unter Verwendung eines genetischen Algorithmus (GA) identifiziert.

Ergebnisse: Das vorgestellte System wurde mittels 1300 benigner Bilder von 10 Patienten mit gutartigen Erkrankungen und 1300 malignen Bildern von 10 Patienten mit Karzinomen überprüft. Wir verwendeten 23 statistisch signifikante (p < 0,0001) Marker. Bei Bewertung des Klassifikators mittels 10-facher Kreuzvalidierung konnten wir eine mittlere Klassifikationsgenauigkeit von 99,8 %, eine Sensitivität von 99,2 % und eine Spezifität von 99,6 % mit einem σ von 0,264 erzielen.

Schlussfolgerung: Das vorgestellte System ist aufgrund der Automatisierung objektiver, es kann in jedem Computer einfach angewandt werden. Es ist schnell und genau und kann den Arzt als ergänzende Methode bei der sicheren Charakterisierung des untersuchten Eierstocktumors unterstützen.

Key words

cervix - 3 D ultrasound - neural networks

• References

• 1 Jemal A, Siegel R, Ward E. Cancer statistics, 2010. CA Cancer J Clin 2010; 60: 277-300
• 2 NIH Consensus Development Panel on Ovarian Cancer. NIH consensus conference. Ovarian cancer. Screening, treatment, and follow-up. JAMA 1995; 273: 491-497
• 3 Horner MJ, Ries LAG, Krapcho M et al. (eds) SEER cancer statistics review, 1975–2006, National Cancer Institute. SEER Website. seer.cancer.gov/csr/1975_2006 Based on November 2008 SEER data submission. Published May 29, 2009.
• 4 Predanic M, Vlahos N, Pennisi JA et al. Color and pulsed Doppler sonography, gray-scale imaging, and serum CA 125 in the assessment of adnexal disease. Obstet Gynecol 1996; 88: 283-288
• 5 Wu CC, Lee CN, Chen TM et al. Factors contributing to the accuracy in diagnosing ovarian malignancy by color Doppler ultrasound. Obstet Gynecol 1994; 84: 605-608
• 6 Iyer VR, Lee SI. MRI, CT, and PET/CT for ovarian cancer detection and adnexal lesion characterization. Am J Roentgenol 2010; 194: 311-321
• 7 Sohaib SA, Reznek RH. MR imaging in ovarian cancer. Cancer Imaging 2007; 7 Spec No A: S119-S129
• 8 Frangioni JV. New technologies for human cancer imaging. J Clin Oncol 2008; 26: 4012-4021
• 9 Anderiesz C, Quinn MA. Screening for ovarian cancer. Med J Aust 2003; 178: 655-656
• 10 Jeong YY, Outwater EK, Kang HK. Imaging evaluation of ovarian masses. Radiographics 2000; 20: 1445-1470
• 11 Barua A, Bitterman P, Bahr JM et al. Contrast-Enhanced Sonography Depicts Spontaneous Ovarian Cancer at Early Stages in a Preclinical Animal Model. Journal of Ultrasound in Medicine 2011; 30: 333-345
• 12 Zhan WW, Zhou P, Zhou JQ et al. Differences in Sonographic Features of Papillary Thyroid Carcinoma Between Neck Lymph Node Metastatic and Nonmetastatic Groups. Journal of Ultrasound in Medicine 2012; 31: 915-920
• 13 Pascual MA, Graupera B, Hereter L et al. Intra-and interobserver variability of 2D and 3D transvaginal sonography in the diagnosis of benign versus malignant adnexal masses. J Clin Ultrasound 2011; 39: 316-321
• 14 Guerriero S, Alcazar JL, Pascual MA et al. Intraobserver and interobserver agreement of greyscale typical ultrasonographic patterns for the diagnosis of ovarian cancer. Ultrasound Med Biol 2008; 34: 1711-1716
• 15 Kim KA, Park CM, Lee JH et al. Benign ovarian tumors with solid and cystic components that mimic malignancy. Am J Roentgenol 2004; 182: 1259-1265
• 16 Acharya UR, Vinitha Sree S, Krishnan MM et al. Non-invasive automated 3D thyroid lesion classification in ultrasound: a class of ThyroScan™ systems. Ultrasonics 2012; 52: 508-520
• 17 Saba L, Gao H, Acharya UR et al. Analysis of carotid artery plaque and wall boundaries on CT images by using a semi-automatic method based on level set model. Neuroradiology 2012;
• 18 Renz C, Rajapakse JC, Razvi K et al. Ovarian cancer classification with missing data. Proceedings of 9th International Conference on Neural Information Processing 2002; 2: 809-813
• 19 Assareh A, Moradi MH. Extracting efficient fuzzy if-then rules from mass spectra of blood samples to early diagnosis of ovarian cancer. IEEE Symposium on Computational Intelligence and Bioinformatics and Computational Biology 2007; 502-506
• 20 Tan TZ, Quek C, Ng GS et al. Ovarian cancer diagnosis with complementary learning fuzzy neural network. Artif Intell Med 2008; 43: 207-222
• 21 Meng H, Hong W, Song J et al. Feature extraction and analysis of ovarian cancer proteomic mass spectra. In: 2nd International Conference on Bioinformatics and Biomedical Engineering 2008; 668-671
• 22 Tang KL, Li TH, Xiong WW et al. Ovarian cancer classification based on dimensionality reduction for SELDI-TOF data. BMC Bioinformatics 2010; 11: 109
• 23 Petricoin F. Use of proteomic patterns serum to identify ovarian cancer. The Lancet 2002; 359: 572-577
• 24 Tailor A, Jurkovic D, Bourne TH et al. Sonographic prediction of malignancy in adnexal masses using an artificial neural network. Br J Obstet Gynaecol 1999; 106: 21-30
• 25 Brüning J, Becker R, Entezami M et al. Knowledge-based system ADNEXPERT to assist the sonographic diagnosis of adnexal tumors. Methods Inf Med 1997; 36: 201-206
• 26 Biagiotti R, Desii C, Vanzi E et al. Predicting Ovarian Malignancy: Application of artificial neural networks to transvaginal and color doppler flow US. Radiology 1999; 210: 399-403
• 27 Zimmer Y, Tepper R, Akselrod S. An automatic approach for morphological analysis and malignancy evaluation of ovarian masses using B-scans. Ultrasound Med Biol 2003; 29: 1561-1570
• 28 Lucidarme O, Akakpo JP, Granberg S et al. A new computer-aided diagnostic tool for non-invasive characterisation of malignant ovarian masses: results of a multicentre validation study. Eur Radiol 2010; 20: 1822-1830
• 29 Bellman RE. Dynamic Programming. Courier Dover Publications; 2003
• 30 Hata T, Yanagihara T, Hayashi K et al. Three-dimensional ultrasonographic evaluation of ovarian tumours: a preliminary study. Hum Reprod 1999; 14: 858-861
• 31 Laban M, Metawee H, Elyan A et al. Three-dimensional ultrasound and three-dimensional power Doppler in the assessment of ovarian tumors. Int J Gynaecol Obstet 2007; 99: 201-205
• 32 Cohen LS, Escobar PF, Scharm C et al. Three-dimensional power Doppler ultrasound improves the diagnostic accuracy for ovarian cancer prediction. Gynecol Oncol 2001; 82: 40-48
• 33 Hu M. Visual pattern recognition by moment invariants. IRE Trans Info Theory 1962; 8: 179-187
• 34 Shen L, Bai L. A review of Gabor wavelets for face recognition. Patt Anal Appl 2006; 9: 273-292
• 35 Manjunath BS, Ma WY. Texture features for browsing and retrieval of image data. IEEE Trans Pattern Anal Mach Intell 1996; 18: 837-842
• 36 Pharwaha APS, Singh B. Shannon and non-shannon measures of entropy for statistical texture feature extraction in digitized mammograms. Proceedings of the World Congress on Engineering and Computer Science 2009; 3: 1-6
• 37 Box JF. Guinness, gosset, fisher, and small samples. Statist Sci 1987; 2: 45-52
• 38 Specht DF. Probabilistic Neural Networks. Neural Networks 1990; 3: 109-118
• 39 Raghu PP, Yegnanarayana B. Supervised texture classification using a probabilistic neural network and constraint satisfaction model. IEEE Trans Neural Netw 1998; 9: 516-522
• 40 Ng EYK, Acharya UR, Keith LG et al. Detection and differentiation of breast cancer using neural classifiers with first warning thermal sensors. Inform Sciences 2007; 177: 4526-4538
• 41 Goldberg DE. Genetic Algorithms in Search, Optimization, and Machine Learning. Addison Wesley Professional Publishers; 1989
• 42 Deb K. Multi-Objective Optimization Using Evolutionary Algorithms. Wiley; 2009
• 43 Bast Jr RC, Badgwell D, Lu Z et al. New tumor markers: CA-125 and beyond. Int J Gynecol Cancer 2005; 15: 274-281
• 44 Zaidi SI. Fifty years of progress in gynecologic ultrasound. Int J Gynaecol Obstet 2007; 99: 195-197
• 45 Menon U, Talaat A, Rosenthal AN et al. Performance of ultrasound as a second line test to serum CA-125 in ovarian cancer screening. BJOG 2000; 107: 165-169
• 46 Acharya UR, Sree SV, Krishnan MRM et al. Ovarian tumor characterization using 3D ultrasound. Technol Cancer Res Treat 2012; (In Press).