Deteksi Kanker Payudara Pada Citra USG Berbasis CLAHE Spring Charged Particle Model (C – SCPM) ANN

Dhaneswar, Tiara Aryacitra (2026) Deteksi Kanker Payudara Pada Citra USG Berbasis CLAHE Spring Charged Particle Model (C – SCPM) ANN. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221053-Ungergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Kanker payudara hingga saat ini masih menjadi salah satu penyebab utama risiko kesakitan dan kematian pada populasi wanita di seluruh dunia, menjadikan strategi deteksi dini sebagai kunci utama dalam upaya penurunan angka kematian. Dalam praktik klinis, Ultrasonografi merupakan salah satu modalitas pencitraan pilihan, terutama bagi wanita dengan jaringan payudara padat, karena sifatnya yang aman, real-time, dan biaya yang relatif ekonomis, dibandingkan dengan MRI. Meskipun demikian, efektivitas diagnostik USG sering kali terhambat oleh kualitas citra yang rendah akibat speckle noise, dan rendahnya kontras antara jaringan lesi dengan jaringan di sekitarnya. Kondisi ini menyebabkan proses penentuan batas massa tumor secara manual menjadi sangat subjektif. Ketergantungan pada interpretasi visual semata dapat berisiko pada kesalahan diagnosis, baik positif palsu maupun negatif palsu. Solusi yang diajukan dalam penelitian ini adalah suatu kerangka kerja terintegrasi yang menggabungkan metode perbaikan kualitas citra dan teknik segmentasi berbasis model deformasi. Pertama, teknik preprocessing menggunakan Contrast Limited Adaptive Histogram Equalization (CLAHE). Selanjutnya Spring Charged Particle Model (SCPM) diimplementasikan untuk melakukan segmentasi lesi. Modifikasi pada model ini dirancang khusus agar sistem partikel memiliki fleksibilitas tinggi dalam beradaptasi terhadap bentuk morfologi tumor ganas yang sering kali tidak beraturan, bersudut tajam, atau memiliki pola spiculated, yang selama ini sulit ditangani oleh metode segmentasi konvensional yang cenderung menghasilkan bentuk yang terlalu halus atau membulat. Dengan ini diharapkan terciptanya suatu sistem deteksi cerdas yang mampu memisahkan area tumor dari jaringan latar belakang dengan presisi dan akurasi yang tinggi. Pada akhirnya, metode ini diproyeksikan dapat berfungsi sebagai perangkat pendukung keputusan (Computer-Aided Diagnosis/CAD) untuk radiolog, meningkatkan efisiensi waktu diagnosis. Hasil pengujian menunjukkan bahwa segmentasi SCPM mencapai koefisien Dice rata-rata 0,8858 dengan waktu komputasi medan gaya 7.000 kali lebih cepat melalui konvolusi 2D. Klasifikasi ANN dengan enam fitur optimal hasil Sequential Forward Selection mencapai akurasi 70,48% pada validasi silang 5-fold.
==================================================================================================================================
Breast cancer remains a leading cause of morbidity and mortality among women worldwide, making early detection an essential strategy for reducing mortality rates. In clinical practice, ultrasonography (US) has become the preferred imaging modality, particularly for women with dense breast tissue, due to its safety, real-time capability, and relatively low cost compared with magnetic resonance imaging (MRI). However, the diagnostic effectiveness of US is often limited by poor image quality, including speckle noise and low contrast between lesion tissue and surrounding tissue. These conditions make the manual delineation of tumor boundaries highly subjective. Reliance solely on visual interpretation increases the risk of diagnostic errors, including both false positives and false negatives. This study proposes an integrated framework that combines image quality enhancement and deformation-model-based segmentation techniques. First, image preprocessing is performed using Contrast Limited Adaptive Histogram Equalization (CLAHE). Subsequently, the Spring Charged Particle Model (SCPM) is applied for lesion segmentation. This model is specifically designed to provide high flexibility in adapting to the morphology of malignant tumors, which often exhibit irregular, angular, or spiculated shapes that are difficult to segment using conventional methods, as these tend to produce overly smooth or rounded boundaries. The proposed framework aims to develop an intelligent detection system capable of isolating tumor regions from surrounding tissue with high precision and accuracy. The system is expected to provide a reliable foundation for the extraction of quantitative morphological features and to serve as a Computer-Aided Diagnosis (CAD) tool for radiologists, thereby improving diagnostic efficiency. The experimental results demonstrate that SCPM segmentation achieves a mean Dice coefficient of 0.8858, while force-field computation is accelerated by approximately 7,000 times through the use of two-dimensional convolution. Furthermore, Artificial Neural Network (ANN) classification using six optimal features selected through Sequential Forward Selection achieves an overall accuracy of 70.48% under stratified five-fold cross-validation.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kanker payudara, Ultrasonografi, Contrast Limited Adaptive Histogram Equalization (CLAHE), Spring Charged Particle Model (SCPM), Segmentasi Citra Medis,Breast Cancer, Ultrasonography, Contrast Limited Adaptive Histogram Equalization (CLAHE), Spring Charged Particle Model (SCPM), Medical Image Segmentation.
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Tiara Aryacitra Dhaneswari
Date Deposited:	31 Jul 2026 10:27
Last Modified:	31 Jul 2026 10:27
URI:	http://repository.its.ac.id/id/eprint/140677

Actions (login required)

View Item