A Methodology for Early Diagnosis of Lung Diseases Based on Multimodal Analysis of Audio Signals and X-Ray Images
Ochiq kirish: yoʻq
Annotatsiya
This paper proposes a methodology for the early diagnosis of lung diseases based on the joint multimodal analysis of respiratory audio signals and chest X-ray images using deep neural networks with an adaptive attention fusion mechanism. The methodology was developed and evaluated on a dataset of 1,240 chest X-ray images and 980 respiratory audio recordings collected from 1,120 patients (aged 18-75 years) at medical institutions in Tashkent, Uzbekistan. On the held-out test set, the proposed multimodal model achieved 95.1% accuracy, F1-score 0.94, outperforming the audio-only (87.9%, F1 0.86) and imageonly (90.3%, F1 0.89) baselines. The approach also enables personalized monitoring of chronic lung conditions such as chronic obstructive pulmonary disease (COPD) and bronchial asthma by comparing new patient data with an individual baseline profile, supporting earlier preliminary diagnosis and remote follow-up.
Hali tarjima qilinmagan