LungVision: Classifying lung diseases using CNN ensembles
Year: 2026
Type: conference-paper
Abstract: COVID-19, viral pneumonia, bacterial infections and other respiratory diseases, represent a major global health problem. The primary, non-invasive tool for diagnosis for these are Chest X-rays (CXRs), but their interpretation by radiologists is time-taking and subjective. The current study proposes a robust deep learning (DL) framework, named LungVision that utilizes a weighted ensemble (weighted) of Convolutional Neural Networks (CNNs) for the multi-class classification of chest X-ray images into four major types: Covid-19, Viral Pneumonia, Normal and few other bacterial infections. The ensemble uses the advance feature extraction capabilities of pre-trained DensNet-121 and ResNet-50 models to increase accuracy and reliability. We utilized Kaggle’s COVID-19 Radiography Dataset, applying minimal to no preprocessing and data modification to enable model predictions on actual clinical data. The proposed work shows better performance in comparison to base models, achieving better result metrics. Experimental results indicate that LungVision is a promising tool for assisting radiologists by providing fast, objective, and reproducible interpretations, potentially accelerating diagnosis in clinical settings.
Institution: ABES Engineering College
Language: English
Topic: COVID-19 diagnosis using AI
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7204203416