SEGMENTASI CITRA DENGAN METODE ATTENTION U-NET UNTUK IDENTIFIKASI LESI PNEUMONIA PADA CITRA RADIOGRAFI

Wulan, Febyu Harnedi (2024) SEGMENTASI CITRA DENGAN METODE ATTENTION U-NET UNTUK IDENTIFIKASI LESI PNEUMONIA PADA CITRA RADIOGRAFI. Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text COVER.pdf Download (1MB)
	Text ABSTRACT (1).pdf Download (41kB)
	Text ABSTRAK.pdf Download (212kB)
	Text BAB 1.pdf Download (129kB)
	Text BAB 2.pdf Download (377kB)
	Text BAB 3.pdf Download (193kB)
	Text BAB 4.pdf Restricted to Registered users only Download (298kB)
	Text BAB 5.pdf Download (59kB)
	Text DAFPUS.pdf Download (218kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (443kB)

Abstract

Pneumonia is a serious lung disease that accounts for 14% of deaths among children under the age of 5, with the death toll reaching approximately 740,180 according to WHO data in 2019. Manual identification of pneumonia lesions on X-ray images can take from several minutes to over an hour per image, thus delaying prompt medical intervention. This study aims to develop a deep learning model using the U-Net architecture that can quickly and accurately segment images to automate the identification of pneumonia lesions on X-ray images, thereby improving the efficiency and accuracy of diagnosis and medical treatment. The U-Net model is trained using a dataset of X-ray images annotated by radiology experts. The image pixel size used is 64 x 64 to ensure consistency in data processing. The model was trained with two different data sets: one set consisting of 20 images over 10 epochs, and the other set consisting of 118 images over 50 epochs. The training results show that a larger dataset and more epochs produce a more accurate model. The trained U-Net model demonstrated good performance with a test loss of 0.6798, a test accuracy of 78.12%, a test precision of 70%, and a test recall of 93.33%. U-Net can accurately identify positive pneumonia cases, thus enhancing the diagnostic process and aiding doctors in providing faster and more accurate treatment. The use of U-Net in identifying pneumonia lesions on X-ray images offers a quick and accurate automated solution, supporting medical personnel in delivering more effective and efficient care.. Keywords: Pneumonia, Image Segmentation, U-Net, Deep learning, Model Accuracy

Actions (login required)

View Item