<![CDATA[Abstract: Lung disease classification based on Chest X-Ray (CXR) images has become an important focus in the development of deep learning for medical imaging. However, most modern classification models still rely heavily on large amounts of labeled data, while medical image annotation requires radiology experts, high costs, and considerable time. This study aims to implement a self-supervised learning approach based on contrastive learning for lung disease classification on CXR images using the CheXpert dataset under limited labeled data conditions. The research stages include data preprocessing, image augmentation, self-supervised pretraining, fine-tuning, and model evaluation using accuracy, precision, recall, and F1-score metrics. The dataset was divided into 70% training data and 30% testing data. The experimental results showed that the model achieved an accuracy of 0.87, precision of 0.87, recall of 0.84, and F1-score of 0.85. These results indicate that the self-supervised learning approach is capable of utilizing unlabeled data to generate robust visual representations and improve lung disease classification performance under limited labeled data conditions. This study is expected to contribute to the development of more efficient deep learning-based medical image analysis systems with reduced dependency on medical annotations. Keywords: Chest X-Ray, Self-Supervised Learning, Contrastive Learning, Deep Learning, Lung Disease Classification. Abstrak: Klasifikasi penyakit paru berbasis citra CXR menjadi salah satu fokus penting dalam pengembangan deep learning pada bidang pencitraan medis. Namun, sebagian besar model klasifikasi modern masih bergantung pada data berlabel dalam jumlah besar, sedangkan proses anotasi citra medis membutuhkan tenaga ahli radiologi, biaya tinggi, dan waktu yang panjang. Penelitian ini bertujuan menerapkan pendekatan self-supervised learning berbasis contrastive learning untuk klasifikasi penyakit paru pada citra CXR menggunakan dataset CheXpert dengan kondisi data berlabel terbatas. Tahapan penelitian meliputi pra-pemrosesan data, augmentasi citra, self-supervised pretraining, fine-tuning, dan evaluasi model menggunakan metrik accuracy, precision, recall, dan F1-score. Dataset dibagi menggunakan rasio 70% data pelatihan dan 30% data pengujian. Hasil penelitian menunjukkan bahwa model mampu menghasilkan nilai accuracy sebesar 0.87, precision 0.87, recall 0.84, dan F1-score 0.85. Hasil tersebut menunjukkan bahwa pendekatan self-supervised learning mampu memanfaatkan data tidak berlabel untuk menghasilkan representasi visual yang robust dan meningkatkan performa klasifikasi penyakit paru pada kondisi data berlabel terbatas. Penelitian ini diharapkan dapat mendukung pengembangan sistem analisis citra medis berbasis deep learning yang lebih efisien terhadap kebutuhan anotasi medis. Kata Kunci: Chest X-Ray, Self-Supervised Learning, Contrastive Learning, Deep Learning, Klasifikasi Penyakit Paru.]]>
