Early detection of tuberculosis using hybrid feature descriptors and deep learning network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Tuberculosis (TB) remains a significant global health challenge, necessitating advanced diagnostic tools.
• Early detection of TB is crucial for effective treatment and disease control.
• Analyzing chest X-ray images is a primary method for TB screening.

Purpose of the Study:

• To develop and evaluate a deep neural network model for early tuberculosis detection from chest X-ray images.
• To compare the efficacy of the proposed deep learning model against existing diagnostic methods.
• To enhance TB diagnosis by integrating deep and hand-engineered features.

Main Methods:

• Utilized an open-source dataset of 4,200 chest X-ray images (3,500 normal, 700 TB).
• Developed a deep learning model combining deep and hand-engineered features for TB diagnosis.
• Applied Gabor filters and Canny edge detection to improve model performance and reduce computational cost.

Main Results:

• The model achieved 95.7% accuracy without image processing techniques.
• Incorporating Gabor filters and Canny edge detection boosted accuracy to 97.9%.
• The model demonstrated superior performance compared to other available models in real-time image testing.

Conclusions:

• The proposed deep learning model offers a promising tool for early TB detection, especially in resource-limited settings.
• The integration of feature engineering techniques significantly enhances diagnostic accuracy.
• The model's effectiveness in real-time scenarios suggests its potential clinical utility.