A Novel and Robust Approach to Detect Tuberculosis Using Transfer Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Disease Research

Background:

• Tuberculosis (TB) remains a global health challenge requiring advanced diagnostic tools.
• Deep learning shows potential in medical image analysis for infectious disease detection.

Purpose of the Study:

• To evaluate the generalizability of deep learning models for tuberculosis detection using chest X-rays.
• To compare the performance of different deep convolutional neural network (CNN) architectures for TB classification.

Main Methods:

• Utilized a publicly available tuberculosis dataset for training and validation.
• Employed image preprocessing, data augmentation, and transfer learning techniques.
• Trained and evaluated four distinct deep CNN models: Xception, InceptionV3, InceptionResNetV2, and MobileNetV2.

Main Results:

• The InceptionResNetV2 model achieved the highest accuracy with an F1-score of 99%.
• The proposed deep learning approach demonstrated superior reliability and accuracy compared to existing research.
• All evaluated CNN models successfully classified tuberculosis and non-tuberculosis cases.

Conclusions:

• Deep learning, particularly the InceptionResNetV2 model, offers a highly accurate and reliable method for computer-assisted TB detection from chest X-rays.
• The developed approach shows promise for rapid and efficient screening of tuberculosis.
• This study highlights the potential of AI in enhancing infectious disease diagnostics.