<![CDATA[ABSTRAKAbstrak— Radiograf toraks (CXR) merupakan alat penting dalam diagnosis penyakit paru, namun interpretasinya memerlukan keahlian khusus dan berpotensi menimbulkan bias. Penelitian ini bertujuan mengeksplorasi kinerja lima arsitektur Convolutional Neural Network (CNN) berbasis transfer learning, yaitu VGG16, ResNet50, EfficientNetB0, DenseNet121, dan MobileNetV2, dalam mengklasifikasikan lima kelas penyakit paru-paru: bacterial pneumonia, COVID-19, tuberculosis, viral pneumonia, dan normal. Dataset yang digunakan dilengkapi dengan preprocessing CLAHE-RGB, augmentasi data, serta penanganan ketidakseimbangan kelas menggunakan class weighting. Evaluasi dilakukan dengan empat skenario epoch (5, 10, 15, dan 30), serta menggunakan metrik akurasi, precision, recall, F1-score, dan confusion matrix. Hasil menunjukkan bahwa model VGG16 pada epoch ke-15 memberikan performa terbaik dengan akurasi 93,95% dan F1-score 0,94. Penelitian ini menunjukkan bahwa kombinasi preprocessing yang tepat dan arsitektur CNN yang sesuai mampu meningkatkan akurasi klasifikasi penyakit paru secara signifikan. Kata Kunci— Convolutional Neural Network, Citra CXR, VGG16, Transfer Learning, CLAHE, Penyakit Paru. ABSTRACTAbstract— Chest radiography (CXR) is a vital tool in diagnosing pulmonary diseases, yet its interpretation often requires expert analysis and may involve subjectivity. This study explores the performance of five Convolutional Neural Network (CNN) architectures: VGG16, ResNet50, EfficientNetB0, DenseNet121, and MobileNetV2 for classifying five categories of lung conditions: bacterial pneumonia, COVID-19, tuberculosis, viral pneumonia, and normal. The dataset underwent preprocessing using CLAHE-RGB enhancement, data augmentation, and class balancing with class weighting. Each model was trained using four epoch scenarios (5, 10, 15, and 30) and evaluated based on accuracy, precision, recall, F1-score, and confusion matrix. The results indicate that VGG16 with 15 epochs achieved the best performance, reaching 93.95% accuracy and 0.94 F1-score. This study demonstrates that combining appropriate preprocessing techniques with suitable CNN architectures significantly enhances classification performance for pulmonary disease detection. Keywords— Convolutional Neural Network, CXR images, VGG16, Transfer Learning, CLAHE, Lung Disease.]]>
