Klasifikasi Kanker Kulit Dari Citra Dermatoskopi Berbasis Convolutional Neural Network U-Net Dan Support Vector Machine

Junior, Amanda Sharon Purwanti (2024) Klasifikasi Kanker Kulit Dari Citra Dermatoskopi Berbasis Convolutional Neural Network U-Net Dan Support Vector Machine. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023201044-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Global Burden of Cancer Study (Globocan) dari World Health Organization (WHO), mencatat jumlah kasus kanker di Indonesia pada tahun 2020 mencapai 396.914 kasus, sementara total kematian akibat kanker mencapai 234.511 kasus. 5,9 – 7,8% dari total kasus kanker yang terjadi merupakan kanker kulit. Tingkat kesembuhan dari kanker kulit dapat meningkat hingga 90% jika ditemukan sejak dini, namun deteksi dini dinilai cukup kompleks dan cenderung subjektif sehingga diagnosis kanker kulit ini seringkali mengalami keterlambatan. Maka dari itu, mulai dikembangkan Computer-Aided Diagnostic (CAD), sebuah sistem diagnosis otomatis yang dirancang dengan tujuan meningkatkan akurasi diagnosis. Diagnosis otomatis pada citra dermatoskopi masih terhambat oleh variasi kompleks dalam tampilan. Dalam penelitian ini, diusulkan sistem yang terdiri dari preprocessing citra yang dilakukan untuk meningkatkan kualitas citra, segmentasi citra menggunakan U-Net yang dilakukan untuk memisahkan lesi dari latar citra, ekstraksi fitur menggunakan metode GLCM untuk menghitung kontras, energi, homogeneity, dan entropi citra pada sudut 0⁰, 45⁰, 90⁰, dan 135⁰ serta metode ABCD yang mengambil beberapa fitur bentuk dan warna pada citra yaitu asimetri(A), tepian atau border (B), warna atau colour (C), dan diameter(D). Terakhir dilakukan klasifikasi multilabel menggunakan Support Vector Machine dengan pendekatan One-Vs-Rest. Model dengan hasil terbaik dalam penelitian didapatkan dengan metode penyeimbangan data SMOTEENN yang merupakan gabungan dari metode SMOTE (Synthetic Minority Over-sampling Technique) dan ENN (Edited Nearest Neighbors) dengan penggunaan kernel Radial Basis Function (RBF) parameter C dan Gamma sebesar 1000 dan 0,1 dengan memakai 21 fitur. Hasil yang didapatkan dari model ini adalah nilai akurasi, presisi, sensitivitas, spesifisitas, dan MCC (Matthews Correlation Coefficient) sebesar 95,25%, 95,25%, 95,24%, 99,02%, dan 0,94.
============================================================
The Global Burden of Cancer Study (Globocan) by the World Health Organization (WHO) reported that the number of cancer cases in Indonesia in 2020 reached 396,914, with total cancer-related deaths reaching 234,511 cases. Skin cancer accounted for 5.9 - 7.8% of the total cancer cases. The cure rate can increase up to 90% with early detection, but early detection is considered complex and subjective, often leading to delayed skin cancer diagnosis. Consequently, a Computer-Aided Diagnostic (CAD) system, designed to enhance diagnostic accuracy, has been developed. Automated diagnosis of dermatoscopic images faces challenges due to complex variations in appearance. In this study, a system is proposed consisting of image preprocessing which is done to improve image quality, image segmentation using U-Net which is done to separate the lesion from the image background, feature extraction using the GLCM method to calculate contrast, energy, homogeneity, and entropy of the image at angles of 0⁰, 45⁰, 90⁰, and 135⁰ and the ABCD method which takes several shape and color features in the image, namely asymmetry (A), edge or border (B), color or color (C), and diameter (D). Finally, multilabel classification is performed using Support Vector Machine with One-Vs-Rest approach. The model with the best results in the study was obtained with the SMOTEENN data balancing method which is a combination of the SMOTE (Synthetic Minority Over-sampling Technique) and ENN (Edited Nearest Neighbors) methods with the use of Radial Basis Function (RBF) kernels with C and Gamma parameters of 1000 and 0,1 using 21 features. The results obtained from this model are the values of accuracy, precision, sensitivity, specificity, and MCC (Matthews Correlation Coefficient) of 95,25%, 95,25%, 95,24%, 99,02%, and 0,94.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kanker Kulit, Segmentasi, Ekstraksi Fitur, Metode ABCD, Metode GLCM, Klasifikasi, CNN, U-Net, SVM, Skin Cancer, Segmentation, Feature Extraction, ABCD Method, GLCM Method, Classification, CNN, U-Net, SVM.
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Amanda Sharon Purwanti Junior
Date Deposited:	05 Aug 2024 01:34
Last Modified:	05 Aug 2024 01:34
URI:	http://repository.its.ac.id/id/eprint/111710

Actions (login required)

View Item