Classification of CT Scan Images of Lungs Using Deep Convolutional Neural Network with External Shape-Based Features

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Accurate lung image classification is crucial for diagnosing respiratory conditions.
• Deep convolutional neural networks (CNNs) show promise in medical image analysis.
• Integrating external features can enhance CNN performance in biomedical tasks.

Purpose of the Study:

• To develop a simplified and efficient deep convolutional neural network (CNN) for lung image classification.
• To incorporate six external shape-based features into the CNN architecture.
• To evaluate the performance of the proposed CNN against existing methods.

Main Methods:

• Utilized computed tomography (CT) scan images from two public databases.
• Developed a deep convolutional neural network (CNN) architecture.
• Embedded shape-based features: solidity, circularity, discrete Fourier transform of radial length (RL), histogram of oriented gradient (HOG), moment, and histogram of active contour image.
• Compared performance using average precision and average recall metrics.

Main Results:

• The proposed CNN achieved an average precision of 95.26%.
• The system obtained an average recall of 69.56% across two databases.
• Performance was benchmarked against six other biomedical image classification methods.

Conclusions:

• The presented simplified CNN architecture offers an efficient approach for lung image classification.
• The integration of shape-based features significantly contributes to improved classification accuracy.
• This method demonstrates strong potential for enhancing diagnostic capabilities in medical imaging.