Ektraksi Fitur Menggunakan Discrete Wavelet Transform dan Full Neighbor Local Binary Pattern Untuk Klasifikasi Mammogram

Putra, Januar Adi (2016) Ektraksi Fitur Menggunakan Discrete Wavelet Transform dan Full Neighbor Local Binary Pattern Untuk Klasifikasi Mammogram. Masters thesis, Institut Teknonologi Sepuluh Nopember.

[thumbnail of 5115201034-Master-Theses.pdf]
Preview
Text
5115201034-Master-Theses.pdf

Download (2MB) | Preview

Abstract

Saat ini pendeteksian kanker payudara dengan citra mammogram telah banyak dilakukan
dengan memanfaatkan pengolahan citra digital. Tahapan dari proses pendeteksian tersebut
terdiri dari preprocessing, ektraksi fitur, seleksi fitur dan klasifikasi. Tahapan yang memegang
peranan penting untuk menghasilkan sistem deteksi yang akurat adalah tahap ekstraksi fitur.
Terdapat beberapa penelitian yang telah dilakukan sebelumnya dengan mengkombinasikan
berbagai metode untuk ekstraksi fitur, salah satu yang menghasilkan akurasi terbaik adalah
kombinasi wavelet dan local binary pattern.. Saat ini pengembangan algoritma local binary
pattern telah banyak dilakukan, salah satunya adalah neighbor local binary pattern (NLBP).
Metode tersebut memiliki perbedaan pada arah dan distribusi relasi spasial dari pixel. Meski
menghasilkan akurasi yang baik, metode NLBP tersebut memiliki beberapa kelemahan yang
sama dengan local binary pattern tradisional, yakni varian terhadap rotasi dan pada proses
thresholding pixel sensitif terhadap noise.
Pada penelitian ini penulis mengusulkan sebuah metode ektraksi fitur baru yang didasarkan
pada neighbor local binary pattern (NLBP). Metode ini memiliki perbedaan pada arah dan
distribusi relasi spasial dari pixel, dimana perbandingan antar pixel pada proses trhesholding
tidak hanya dengan tetangga di bagian kanan saja melainkan dengan semua tetangga yang ada
pada sisi horizontal, vertical dan diagonal sehingga metode tersebut disebut full neighbor local
binary pattern (FNLBP). Metode ini nantinya akan dikombinasikan dengan discrete wavelet
transform untuk ektraksi fitur dari citra mammogram dengan classifier adalah Backpropagation
Neural Network (BPNN).
Berdasar ujicoba yang telah dilakukan metode usulan mendapatkan rata-rata akurasi yang
lebih baik daripada metode local binary pattern tradisional baik yang dikombinasi dengan
discrete wavelet transform ataupun tidak. Performa metode usulan full neighbor local binary
pattern dapat menghasilkan akurasi yang tinggi yakni 92.70% pada saat menggunakan discrete
wavelet transform dengan seleksi fitur f-test dengan significant level 0.9 dan 0.7, sedangkan
akurasi terendah yang didapat adalah saat digunakan metode seleksi fitur t-test dengan nilai
significant level 0.5 pada kombinasi discrete wavelet transform dan full neighbor local binary
pattern yakni 77.08%.

====================================================================================

Currently the detection of breast cancer with a mammogram image has much to do with
utilizing digital image processing. Stages of the detection process consists of preprocessing,
feature extraction, feature selection and classification. Stages which plays an important role to
produce an accurate detection system is the feature extraction stage. There are several studies
that have been done before by combining various methods for the extraction of features, the one
that produces the best accuracy is a combination of wavelet and local binary pattern. Currently
the development of local binary pattern algorithms have been done, one of which is a neighbor
of local binary pattern (NLBP). Such methods have differences on the direction and distribution
of spatial relationships of pixels. Although it provides good accuracy, NLBP methods have
some disadvantages similar to traditional local binary pattern, which is a variant of the rotation
and the thresholding of the pixels are sensitive to noise.
In this study, the authors propose a new feature extraction method based on local neighbor
binary pattern (NLBP). This method has a difference in the direction and distribution
relationships spatial pixel, where a comparison between the pixels of the process trhesholding
not only with neighbors on the right side only, but with all the neighbors were there on the side
of the horizontal, vertical and diagonal so the method called full neighbor local binary pattern
(FNLBP). This method will be combined with discrete wavelet transform to extract the features
of the mammogram image with a classifier is Backpropagation Neural Network (BPNN).
Based on experiments the result of proposed method in an average accuracy is better than
traditional methods of local binary pattern which combined with discrete wavelet transform or
not. The performance of the proposed method of full neighbor local binary pattern can produce
high accuracy that is 92.70%, this accuracy is reached when using discrete wavelet transform
with selection feature method is f-test and the significant level is 0.9 and 0.7, while the lowest
accuracy obtained are currently when use t-test feature selections method with a value of
significant level is 0.5 on a combination of discrete wavelet transform and full local neighbor
binary pattern that is 77.08%.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Ekstraksi fitur; local binary pattern; wavelet; klasifikasi mammogram; Feature extraction; local binary pattern; mammogram classification
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User: - JANUAR ADI PUTRA
Date Deposited: 04 Apr 2017 01:28
Last Modified: 27 Dec 2018 04:26
URI: http://repository.its.ac.id/id/eprint/2533

Actions (login required)

View Item View Item