Delayed diagnosis continues to negatively impact the treatment of pulmonary diseases. It has become a leading cause of mortality, especially in regions with limited access to healthcare infrastructure. In this paper, we propose an intelligent system for the early detection of pulmonary diseases. Our approach combines two complementary data modalities: chest X-ray images and respiratory audio signals. In this work, we used ResNet50 to identify lung conditions from X-ray images. For the audio signals, we converted them into spectrograms and used ResNet18 for classification. The database is composed of images and audio recordings collected from various sources. The system achieved diagnostic accuracy rates ranging from 90% to 98% across multiple datasets. These results demonstrate that integrating multiple modalities significantly improves the diagnosis of pulmonary diseases. This research proposes a new dual-modality diagnostic system that consists of visual and audio data streams adapted to remote and low resource settings, and aims to contribute to quick, accurate and available diagnoses in real clinical environments.