Related Experiment Video
Updated: Dec 2, 2025
10:26
Author Spotlight: A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
2.3K
A bilinear convolutional neural network for lung nodules classification on CT images
Rekka Mastouri1, Nawres Khlifa2, Henda Neji3,4
1Higher Institute of Medical Technologies of Tunis, Research Laboratory of Biophysics and Medical Technologies, University of Tunis el Manar, 1006, Tunis, Tunisia. rekka.mastouri@gmail.com.
International Journal of Computer Assisted Radiology and Surgery
|November 3, 2020
Summary
A new bilinear convolutional neural network (BCNN) effectively classifies lung nodules in CT scans, improving early lung cancer detection. This AI approach offers promising results for radiologists, enhancing diagnostic accuracy.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Oncology
- Deep Learning for Diagnostics
Background:
- Lung cancer is a leading cause of cancer-related deaths globally.
- Early detection of lung nodules is crucial for improving patient prognosis.
- Computer-aided diagnosis systems show promise in medical image analysis.
Purpose of the Study:
- To introduce a novel Bilinear Convolutional Neural Network (BCNN) for lung nodule classification on CT images.
- To evaluate the performance of BCNN in comparison to existing deep learning architectures.
- To assess the potential of BCNN as a tool for early lung cancer diagnosis.
Main Methods:
- Utilized two-stream Convolutional Neural Networks (CNNs), VGG16 and VGG19, as feature extractors.
- Implemented a Support Vector Machine (SVM) classifier for false positive reduction.
- Experimented with BCNN combinations and various SVMs to determine optimal classification.
Main Results:
- The BCNN [VGG16, VGG19] combination achieved an accuracy of 91.99% and an AUC of 95.9%.
- Performance surpassed individual [VGG16]2 and [VGG19]2 architectures.
- Results were validated on 3186 images from the LUNA16 database.
Conclusions:
- The proposed BCNN method demonstrates improved outcomes over conventional CNN architectures.
- BCNN offers a promising and computationally affordable approach for lung nodule analysis.
- BCNN can serve as a valuable assessment tool for radiologists in early lung cancer diagnosis.
