Klasifikasi Pneumonia Anak Pada Citra X-Ray Dada Menggunakan Metode Adaptive Masking dan VGG-16

Herawati, Yoshi Inne (2024) Klasifikasi Pneumonia Anak Pada Citra X-Ray Dada Menggunakan Metode Adaptive Masking dan VGG-16. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (cover) 20081010127.-Cover.pdf Download (871kB)
	Text (Bab 1) 20081010127.-Bab1.pdf Download (92kB)
	Text (Bab 2) 20081010127.-bab2.pdf Restricted to Repository staff only until 30 October 2026. Download (314kB)
	Text (Bab 3) 20081010127.-bab3.pdf Restricted to Repository staff only until 30 October 2026. Download (578kB)
	Text (Bab 4) 20081010127.-bab4.pdf Restricted to Repository staff only until 30 October 2026. Download (1MB)
	Text (Bab 5) 20081010127.-bab5.pdf Download (16kB)
	Text (Daftar Pustaka) 20081010127.-daftarpustaka.pdf Download (147kB)

Abstract

This study aims to analyze and compare the classification accuracy of childhood pneumonia using the VGG-16 method by applying various pre processing techniques, including histogram equalization, CLAHE, Gaussian Blur, and adaptive masking on chest X-ray images. The three main objectives of this study were to obtain and compare high accuracy values, apply these pre-processing techniques, and evaluate the accuracy results obtained. After a series of test scenarios, it was found that the combination of histogram equalization, Gaussian Blur, and adaptive masking with the use of SGD optimizer achieved the highest accuracy of 87%, with a precision value of 86%, recall of 87%, and F1-score of 86%. Adaptive masking is proven to contribute significantly to improving classification accuracy compared to methods that do not use adaptive masking. In addition, SGD optimizer shows better performance compared to other optimizers such as RMSprop and Adamax. This study also found that the setting of 30 epochs, batch size 32, and learning rate 0.001 is the optimal configuration to achieve stable model convergence.

Item Type:

Thesis (Undergraduate)

Contributors: