Related Experiment Video
Updated: Sep 12, 2025
03:38
Unilateral Lung Volume Analysis Using Micro-CT for Enhanced Assessment of Pulmonary Fibrosis in Preclinical Models
Published on: June 20, 2025
323
RETRACTED: LGD_Net: Capsule network with extreme learning machine for classification of lung diseases using CT scans
Ali Haider Khan1, Jianqiang Li1, Muhammad Nabeel Asghar2
1College of Computer Science, Beijing University of Technology, Beijing, China.
Plos One
|August 8, 2025
Summary
A novel LGD_Net model accurately classifies lung diseases (LGDs) like pneumonia and cancer using CT scans. This AI approach significantly aids radiologists in early and precise LGD diagnosis, improving patient outcomes.
Area of Science:
- Medical Imaging and Artificial Intelligence
- Computational Pathology
- Radiology
Background:
- Lung diseases (LGDs) encompass various conditions like pneumonia, lung cancer, tuberculosis, and COVID-19.
- Similar symptoms and imaging findings across LGDs present diagnostic challenges for radiologists.
- Timely and accurate diagnosis is critical to prevent severe complications and mortality.
Purpose of the Study:
- To propose a novel deep learning model, LGD_Net, for accurate classification of five LGDs using CT scans.
- To enhance LGD classification by combining Capsule Networks (CapsNet) with Extreme Learning Machines (ELM).
- To address data imbalance and improve dataset quality using Borderline-SMOTE and affine transformations.
Main Methods:
- Development of LGD_Net, integrating CapsNet and ELM architectures.
- Training and testing on five public benchmark CT scan datasets.
- Application of Borderline-SMOTE for imbalanced data and affine transformations for data augmentation.
Main Results:
- LGD_Net achieved a superior accuracy of 99.71% in classifying LGDs.
- Performance significantly outperformed established CNN models: Vgg-19 (91.21%), ResNet-101 (94.39%), Inception-v3 (93.96%), and DenseNet-169 (93.82%).
- Demonstrated state-of-the-art (SOTA) performance in LGD classification.
Conclusions:
- The LGD_Net model offers a highly accurate and effective automated solution for LGD classification from CT scans.
- This AI-driven approach provides substantial support to radiologists, enhancing diagnostic capabilities.
- The findings highlight the potential of LGD_Net to improve early detection and management of lung diseases.
