skip to main content

DETEKSI EFUSI PLEURA PADA CITRA THORAX MENGUNAKAN JARINGAN SYARAF TIRUAN PROPAGASI BALIK MELALUI EKSTRAKSI CIRI BINER

*Elvira Situmorang  -  Jurusan Fisika, Fakultas Sains dan Matematika, Universitas Diponegoro, Semarang, Indonesia
Kusworo Adi  -  Jurusan Fisika, Fakultas Sains dan Matematika, Universitas Diponegoro, Semarang, Indonesia
Evi Setiawati  -  Jurusan Fisika, Fakultas Sains dan Matematika, Universitas Diponegoro, Semarang, Indonesia

Citation Format:
Abstract

The research about detection pleural effusion of the thoracic using neural network back propagation by binary feature extraction has been done. A common cause of pleural effusion disease is cancer. It is estimated that pleural effusion malignant affects 150,000 people every year in the United States. The normal pleural space only has a few milliliters of liquid that helps lubricate of the lungs during breathing. Pleural effusion (large amounts of liquid in the pleural space) can lead to a partial or complete compression of the lung. The difficulty to distinguish excess accumulation of fluid in the pleural cavity should be minimized by radiologist. This research contributes interpretation pleural effusion in the thoracic and reduces doubts of doctor in the treatment of patients. The purpose of this research is to develop algorithms to identify pleural effusion using artificial neural networks back propagation by binary feature extraction the thoracic. Binary feature extraction is obtained from the level set segmentation. The process of image enhancement by histogram equalization and contrast enhancement should be performed before the level set segmentation process. Binary feature extraction patterns were training on ANN was taken from 5% until 25% of costophrenic angle in the thoracic. Neural network can recognize the characteristic patterns of the binary feature 15% are well trained. Validation ANN pattern training by up to 100%, while process of testing the ANN is able to identify 14 data from 15 test data to test validation value reaches 93.33% on the condition of  setting 2 hidden layers, each of hidden layer contain 10 neurons.

 

Keywords: Pleural effusion, Binary feature extraction, Artificial neural networks, Histogram, level set segmentation.

Fulltext View|Download

Last update:

No citation recorded.

Last update:

No citation recorded.