COMPARATIVE ANALYSIS OF SKIN CANCER IMAGE CLASSIFICA-TION USING KERNEL-VARIANT SUPPORT VECTOR MACHINES AND CONVOLUTIONAL NEURAL NETWORK-BASED DEEP LEARN-ING APPROACHES

International Journal of Computer Science (IJCS) Published by SK Research Group of Companies (SKRGC)

Abstract

Early detection of skin cancer significantly reduces the risk of developing malignant tumors such as melanoma. Skin tumors can be either benign or malignant, and their classification relies on visible symptoms and characteristic features. This study utiliz-es a clinically validated image dataset containing over 25,000 skin tumor images for classification purposes. From these images, categorical and continuous features were extracted, resulting in a structured dataset of 10,000 images described by eleven at-tributes related to cellular properties. Initially, a Support Vector Machine (SVM) clas-sifier was trained and evaluated on this dataset. The SVM models employed four different kernel functions—Linear, Polynomial, Sigmoid, and Radial Basis Function (RBF)—to map the features into higher-dimensional spaces. Their classification per-formance was assessed using metrics such as accuracy, specificity, and sensitivity. In the second phase of this research, the same dataset was used to develop and evalu-ate deep learning models based on Convolutional Neural Networks (CNNs). Three custom CNN architectures were designed by adding additional convolutional, pool-ing, and hidden layers, and optimizing various parameters for improved feature ex-traction and classification performance. Additionally, pre-trained models such as VGG16 and ResNet50 were employed for comparative analysis. The performance of these CNN-based models was measured and compared against the SVM classifiers to determine the most effective approach for skin cancer image classification.

References

1) Yuan X, Yang Z, Zouridakis G, Mullani N. SVM-based texture classification and application to early melanoma detection. Conf Proc IEEE Eng Med Biol Soc. 2006;2006:4775-8. doi: 10.1109/IEMBS.2006.260056. kaggle website, https://www.kaggle.com/andrewmvd/isic-2019

2) Tajeddin NZ, Asl BM. Melanoma recognition in dermoscopy images using lesion's peripheral region information. Compute Methods Programs Biomed. 2018 Sep;163:143-153. doi: 10.1016/j.cmpb.2018.05.005.

3) Tabatabaie K, Esteki A, Toossi P. Extraction of skin lesion texture features based on independ-ent component analysis. Skin Res Technol. 2009 Nov;15(4):433-9. doi: 10.1111/j.1600-0846.2009.00383.x.

4) Serban ED, Farnetani F, Pellacani G, Constantin MM. Role of In Vivo Reflectance Confocal Microscopy in the Analysis of Melanocytic Lesions. Acta Dermatovenerol Croat. 2018 Apr;26(1):64-67.

5) Efimenko M, Ignatev A, Koshechkin K. Review of medical image recognition technologies to detect melanomas using neural networks. BMC Bioinformatics. 2020 Sep 14;21(Suppl 11):270. doi: 10.1186/s12859-020-03615-1.

6) Harangi B. Skin lesion classification with ensembles of deep convolutional neural networks. J Biomed Inform. 2018 Oct;86:25-32. doi: 10.1016/j.jbi.2018.08.006.

7) Tanaka M, Saito A, Shido K, Fujisawa Y, Yamasaki K, Fujimoto M, Murao K, Ninomiya Y, Satoh S, Shimizu A. Classification of large-scale image database of various skin diseases using deep learning. Int J Comput Assist Radiol Surg. 2021 Jul 26. doi: 10.1007/s11548-021-02440-y.

8) C. K. Bichakjian, A. C. Halpern, T. M. Johnson, A. F. Hood, J. M. Grichnik, S. M. Swetter, H. Tsao, V. H. Barbosa, T. Y. Chuang, M. Duvic, and V. C. Ho, ‘‘Guidelines of care for the management of primary cutaneous melanoma,’’ J. Amer. Acad. Dermatol., vol. 80, no. 1, pp. 208–250, 2019.

9) American Cancer Society. Facts & Figures, Amer. Cancer Soc., Atlanta, Ga, USA, 2020.

10) R. J. Friedman, D. S. Rigel, and A. W. Kopf, ‘‘Early detection of malignant melanoma: The role of physician examination and self-examination of the skin,’’ CA A, Cancer J. Clinicians, vol. 35, no. 3, pp. 130–151, 1985.

11) Z. Apalla, D. Nashan, R. B. Weller, and X. Castellsagué, ‘‘Skin cancer: Epidemiology, dis-ease burden, pathophysiology, diagnosis, and therapeutic approaches,’’ Dermatol. Therapy, vol. 7, no. S1, pp. 5–19, Jan. 2017.

12) Pham, T.-C., Luong, C.-M., Visani, M., Hoang, V.-D.: Deep CNN and Data Augmentation for Skin Lesion Classification, pp. 573–582 (2018). (http://link.springer.com/10.1007/978-3-319-75420-8_54)

13) Author, F., Author, S., Author, T.: Book title. 2nd edn. Publisher, Location (1999).

14) S. Pathan, K. K. Pandey, and M. Raut, “Skin lesion classification using deep learning and transfer learning,” Procedia Computer Science, vol. 132, pp. 879–887, 2018, doi: 10.1016/j.procs.2018.05.146.

15) N. Codella et al., “Skin lesion analysis toward melanoma detection: A challenge at the 2017 ISIC workshop,” in 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018), Washington, DC, USA, 2018, pp. 168–172, doi: 10.1109/ISBI.2018.8363547.

16) M. Tschandl, C. Rosendahl, and P. Kittler, “The HAM10000 dataset: A large collection of multi-source dermatoscopic images of common pigmented skin lesions,” Scientific Data, vol. 5, 180161, 2018, doi: 10.1038/sdata.2018.161.

17) Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, May 2015, doi: 10.1038/nature14539.

18) A. Esteva et al., “Dermatologist-level classification of skin cancer with deep neural net-works,” Nature, vol. 542, no. 7639, pp. 115–118, Feb. 2017, doi: 10.1038/nature21056.

19) H. Abbas, M. Abd El-Rahman, and A. Elmogy, “A convolutional neural network for skin lesion classification using multi-view data,” Multimedia Tools and Applications, vol. 80, pp. 27383–27402, 2021, doi: 10.1007/s11042-021-10791-7.

20) R. P. Tschandl, “Using deep convolutional neural networks for the classification of dermato-scopic images,” IEEE Access, vol. 6, pp. 11482–11493, 2018, doi: 10.1109/ACCESS.2018.2794701.

21) J. Kawahara, A. BenTaieb, and G. Hamarneh, “Deep features to classify skin lesions,” in 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI), Prague, Czech Repub-lic, 2016, pp. 1397–1400, doi: 10.1109/ISBI.2016.7493481.

22) S. Mahbod et al., “Fusing fine-tuned deep features for skin lesion classification,” Computer-ized Medical Imaging and Graphics, vol. 71, pp. 19–29, Jul. 2019, doi: 10.1016/j.compmedimag.2018.10.002.

23) R. Yu et al., “Automated melanoma recognition in dermoscopy images via very deep residual networks,” IEEE Transactions on Medical Imaging, vol. 36, no. 4, pp. 994–1004, Apr. 2017, doi: 10.1109/TMI.2016.2642839.

Keywords

Support Vector Machine, CNN Classifier, VGG16, Diverse Kernels, ResNet50.