AI-Based Multi-Disease Detection Application Using Multimodal Deep Learning

Respiratory diseases remain a leading cause of illness and death worldwide, yet diagnosing them usually means reading a chest X-ray, listening to breath sounds and reviewing clinical notes as three separate exercises.This paper presents a multimodal deep learning application that brings these sources together to screen for asthma, chronic obstructive pulmonary disease, pulmonary fibrosis, pneumonia and tuberculosis.Chest X-rays are analysed with a ResNet50 network, breathing recordings are converted into Mel-frequency cepstral coefficients and classified by a dense neural network, and clinical notes are vectorised with term frequency-inverse document frequency and classified by a second dense network.The three calibrated confidence scores are combined by late weighted fusion, with weights of 45%, 30% and 25% for image, audio and text, and the fused score is mapped to a disease decision and a risk level.Gradientweighted class activation maps show which lung regions influenced the image prediction.The models sit inside a secure web application that stores patient records and produces downloadable reports.The system is built on a balanced dataset of 1,500 patient records, and its end-to-end behaviour is demonstrated on disease-present and normal cases.It is meant to support clinicians, not replace them.